This action might not be possible to undo. Are you sure you want to continue?
ADA280 631
DTIC ELECTE
S
S~G
JUN 2b41994
AN AERODYNAMIC AND STATICSTABILITY ANALYSIS OF THE HYPERSONIC APPLIED RESEARCH TECHNOLOGY (HART) MISSILE DISSERTATION Kenneth John Moran Captain, USAF AFIT/DS/AA/943
9419380
Approved for public release; distribution unlimited
AFIT/DS/AA/943
AN AERODYNAMIC AND STATICSTABILITY ANALYSIS OF THE HYPERSONIC APPLIED RESEARCH TECHNOLOGY (HART) MISSILE
DISSERTATION
Presented to the Faculty of the Graduate School of Engineering of the Air Force Institute of Technology Air University In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy Accesion For NTIS
DTIC
CRA&I
TAB
0 Unannounced Justification.................... Kenneth John Moran, B.S., M.S. Captain, USAF By .................. Distuibution I
Availability Codes Avail and i or Special
=
Dist
April, 1994
Approved for public release; distribution unlimited
AFIT/DS/AA/943
AN AERODYNAMIC AND STATICSTABILITY ANALYSIS OF THE HYPERSONIC APPLIED RESEARCH TECHNOLOGY (HART) MISSILE
Kenneth John Moran, B.S., M.S. Captain, USAF Approved:
J. S. Przemienicki Senior Dean
Acknowledgments
I dedicate this dissertation to my late father. It was always his desire to see me fulfill my dream of becoming an aeronautical scientist. Although he is not here to share this moment, his memory will always give my accomplishments meaning. His efforts and sacrifices made it all possible. I am grateful to many people and several organizations for their assistance during my research. In particular, my advisor Dr. Philip Beran gave me the liberty to pursue my own goals and the guidance to realize them. I am forever in his debt. I would also like to thank the other committee members, Dr. Dennis Quinn, Captain John Doty, Dr. Rodney Bowersox, and my sponsor Dr Bruce Simpson. I am indebted to Dr Simpson on two counts, first for suggesting the research topic and second for providing invaluable computing resources without which this work could not have been done. Likewise, I owe thanks to the Avionics Directorate of Wright Laboratory whose Convex C220 came to reside at AFIT and became a computational work horse. Also, I would like to acknowledge the DoD National High Performance Computing Shared Resource Centers who provided some additional computing resources. My wife, Tracy, deserves special recognition. Her tireless support and limitless understanding gave me sustenance and strength in ways I am only now beginning to realize.
Kenneth John Moran
111II
........ Scope of Thickfin Investigation (Part II).... 2......... 12 12 12 . Abstract ... .... 1... iii iv ix xvi xviii xx 1 1 3 3 5 6 6 8 9 10 List of Figures .................................................. .... . 2..............1.... List of Symbols ........1................ ...... 1..................... Governing Equations and Supporting Theory ............................1 Background on Missile Research .. Overview of Dissertation .................. Introduction .......3 1.......... 1.....1..... 1...................4........2 Scope of Thinfin Investigation (Part I) . Summary of Previous Results ...Table of Contents Page Acknowledgments Table of Contents ...........1............................ List of Tables ....................4 Background on Fin/Body Interaction Problem ................ Analytical and Numerical Testing ............................... Part I: Investigation Using "Infinitely Thin Fin" Approximation 2....4.................. 1................ .. Purpose for Present Research ..1 Shock Wave Shape .................................3 Experimental Testing ..1 Flow Structure Near Fin ............ iv ........1 1....2 1................. .. ..............2 1.......1 1...
.....1 2.............2 Theoretical Forebody Wave Drag .............. Methodology for the Thinfin Investigation ..3........2....6............. 3.......... Empirical Model in Wake ... Computational Approach .......... .........4... Aerodynamic Coefficients ........ 13 15 15 ............. 3.....4. Impermeability Condition for a Threedimensional Surface ... Cd ...... Modification to the Length Scale in the Turbulence Model ............. Pitching Moment From Linear Theory ... 2.... 3........1 3...2.........2 2......... Pitchingmoment Coefficient...... 3................... General Algorithm Description ........................5 Grid Structure .. 16 18 22 22 24 26 26 26 27 27 29 38 38 39 39 40 46 46 47 48 50 Inviscid Theory for Drag and Pitching Moment ... .4.......3 Singularline Conditions ..........2 .....2 Overview of the Development of the Flowfield Solver.2 3......4......2 Shock/Boundarylayer Interaction . 3. ..........5....1 Grid Limitations ....3 Hierarchy of Effects ...................................6 Boundary Conditions ......... Trailingedge...............4 Turbulence Modeling ......... Drag Coefficient....3 2.........1 2... Description of the Thinfin Approximation ..Page 2.....4 Governing Equations and Boundary Conditions ..................................2 3... 3................ 3...........6...............2 3.....3.......... and Surface Conditions ........1 3... Fin Leadingedge.6. v ..1 3........ 3........ 2. 2...1.......1 3. 3..3 Compressibility and Pressuregradient Modifications ..... C.4.............
...... Grid Refinement and Sensitivity ......................3..........4 4. 4........5 4...2 4....... Uncertainty Due to Boundarylayer Transition....... Total Drag....... Results and Analysis from Thinfin Tests .............1 6..4...2 ... 4.. ...1 4........ 4.................4 Drag Analysis ................4................. Comparison Between the Clippeddeltafin Config........... Flow Structure Near a BluntSweptFin/Body Junction ......... 6....... 4. Laminar Results ........3 4... Shockwave Shape . ..... Effect of Fins on Base Drag ....................4.. Influence of Base Flow . ..4.........5 Staticstability Analysis ..6 Forebody Wave Drag..... 4. Shock/Boundarylayer Interaction .......... 51 51 52 57 57 58 61 63 65 67 67 67 70 72 77 82 87 96 101 Summary of Numerical Experiments ......5......1 4....... 106 Part II: Investigation Using Thick Fin 6.....5.......... Summary and Conclusions from Thinfin Investigation ............................5 4.4...... Forebody Friction Drag.......101 uration Lnd the Baseline Configuration .......1 4................ Turbulent Results ..............................3 ........1 4..3..5...5.......... 4.. Base Drag..2 4.......2 4....................3 4... Inviscid Results .4..... Methodology for Thickfin Investigation ..................5..6 Experimental Model Uncertainty..4 4.......... Flowfield Near Fin .....Page 4..... 111 111 112 vi .......... 5.5...2 Increased Modeling Complexity ..........
....4..... ..2 7..............3 7...1 7..1.............1 6.....1 Axisymmetric Equations .......2............ Stability Analysis for Source Term ...........6 Turbulence Modeling ... 123 126 128 130 ...2........... A....1 Reformulation of Finitedifference Equations ........2 Governing Equations for Finitevolume Method....................3 Computations for a Pointed and Blunted Tangent OgiveCylinder (PTOC and BTOC) in Turbulent Flow .... ..... A.. HemisphereCylinder (HC) ....... ..... A.....1...............4 Results and Conclusions from Thickfin Investigation ...3 Suggested Boundary for Incipient Separation at Fin Leading Edge ........... A.......4 A......1... Algorithm Validation Using Axisymmetric Equations A.1..............4 Flow Structure Changes Near Fin .... Grid Structure .... Summary and Conclusions of ThickFin Analysis Appendix A......... Analysis of Fin Drag .Page 6.5 Boundary Conditions .....1............3 Discretization of the NavierStokes Equations................1 A.....1. ... .... 7.2 Integration of the NavierStokes Equations for a Laminar Forebody .... A..2 7..........7 Computer Resources ...... 138 138 145 149 A..........2 A. ..... .... A..... ...3 Spherical Nosetip .........2..... 112 114 114 116 117 117 121 ........ Impact of Fin Interaction on Static Stability . ...... Numerical Implementation ... A.... Highly Blunted Tangent OgiveCylinder (HBTOC)... A.......... 153 vii ..4.... Modification of Boundary Conditions ....1. 6...... 130 132 133 134 135 136 138 6..........
.4 Calculations for a CylinderFlat Base (CFB) Configuration Using an Empirical Turbulence Model .... 164 165 168 170 B.............. .... C.... 174 177 180 187 .......... .................... Appendix C..... 190 193 202 viii .... 172 B....5 Conclusions from Threedimensional Validation ..... 173 B..... C............... ............. Vita ...2 Algorithm Performance and Memory Requirements .......1 Modifications to Algorithm for Threedimensional Equations 170 B.................. Computer Codes and Resources .......2 Application to Axisymmetric Baseline .......... A...5 Convergence Acceleration Techniques for SteadyState Computations ....6 Conclusions from Axisymmetric Validation ............ B...... B.... ......1...............2 Geometric Terms of a Threedimensional Fluid Cell. 171 B........ Appendix B..... Validation of Threedimensional Equations .... Bibliography ..1... .................... A.. 188 ... .. ..4 Calculations for an OgiveCylinder Forebody at Angle of Attack (Turbulent Flow) ..1 Software Documentatiion for General Algorithm ..................3 Nonconformal Grid Applied to Axisymmetric Compression Ramp ....Page A................ ...1 Generalized Jacobian and eigenvector ......... .........
... 2 4 7 7 13 14 14 18 22 30 Shockwave angles measure from the centerline of a delta wing (from ............... Axialforce coefficient (drag) for HART baseline model .. . AO'......... 8... 17..... 43 44 45 49 54 Grid structure near singular line: (a) finitedifference.... 4........... Modified nosetip showing very slight rounding of sharp nosetip ..... Suggested flow patterns for sweptsharpfin/sidewallboundarylayer in. 3....... 0 [At 2 ............ (b) (Re=100.. (c) aftend view ........ Typical fluid cell to which conservation laws are applied ... Baseline grid: (a) surface grid near nosetip... CFL = 0..... AX2] .000. 11...... azimuthal angle................ .... 2. .... Pitchingmoment coefficient for HART baseline model .... 0 [At 2 ..... the fin ........ .. 10.. 14....... C.. 18..... Ax 2]......... Accuracy of scheme when applied to linear viscous Burgers equation (a) current scheme.. Pitching moment using thinairfoil and slenderbody theories Computational domain showing boundary conditions .............. 7....... ix ... Baseline grid cutaway view: 60 x 81 x 33 around forebody and 31 x 111 x 33 42 .. Ax 2].......... Swept thin fin geometry .List of Figures Figure Page Hypersonic Applied Research Technology (HART) baseline model HART models with modified nosetip and clipped delta fins ...... [80] with permission) . .... 1. 5... 4'..... .......... (b) finitevolume Body orientation showing references for body inclination angle........... 6..... 0 [At...000................ 0.... .. 16.. ...... AX2 ] . near .... Accuracy of scheme when applied to linear viscous Burgers equation (a) current scheme. Sensitivity of axialforce coefficient. and ... CFL = 1.... 9.. 15........... 37 13. (b) (Re=100.... in the wake .....0): ATNSC2 scheme......... where y = r cos 4' and z = r sin 4 .... to angular spacing.... teractions (from [80] with permission) .. 0 [At..9): ATNSC2 scheme.... 36 12..
.... between computation and experiment for the sharpnosetip....... HART missile . Viscousdrag coefficient... Cdb..... .... Cdib.... Pressure contours for the fin on the compression side of the missile ....... .. Pitchingmoment coefficient. x 75 76 34............. for bluntnosetip model (a = 0) ....... 24.. showing effect of angle of attack . 28............... clippeddeltafin...... (b) flatplate model .. for the sharpnosetip model in turbulent . clippeddeltafin. for HART missile showing restabilization trend at higher Mach numbers . (Case B8): (a) upper surface.... Comparison of the totaldrag coefficient........ Cd.. Pressure contours on the missile near the fin/body junction (Case B8) Pressure contours for the fin on expansion side of the missile (Case B8): (a) upper surface.. AO)...... C............ 32.. CdQ.................................. 21.. .. Cdi.................. Comparison of the totaldrag coefficient. clippeddeltafin....... (b) lower surface ....... 74 33. .. HART missile (a = 00): (a) fin friction drag not accounted for. Cd........ near the fin . Cd.... Basedrag coefficient........ Page 54 57 58 59 60 20. (b)turbulent flow (Case S28) ... extracted from freeflight tests of the sharpnosetip......... Sensitivity of pitchingmoment coefficient.. 78 .. (b) lower surface .... (a=0 Pitchingmoment coefficient......... 35. HART missile 0 ) .............. Pressure along the missile body upstream of the fin/body junction: (a) laminar flow (Case S16). 69 69 73 Basedrag coefficient... between computation and experiment for the bluntnosetip. to angular spacing............. Wavedrag coefficient..... C...... 62 66 66 68 68 25....... Totaldrag coefficient..... 26.... 23..... Cd....... 29......... showing fin interference effects (a = 0*) .. Basedrag coefficient............ Sensitivity of pitchingmoment coefficient. Wavedrag coefficient....... 27. for sharpnosetip model (a = 00) . C... for HART missile from reevaluated ........ freeflight data .....Figure 19..... C..•. Cd. 31..... to angle of attack. 22.... 30.. .. flow (a = 00 ) .. for the HART missile (a = 00) . of fin friction drag .....
. Pressure contours on the fin surfaces for Case S16 (M.. 51.o = 3......... = 2... 37........5... = 2.... Cm... (b) inverteddeltafin models .. a = 50.. Documented fin size and position of clippeddelta fin (normalized by ....... Cm... Pressure contours.. from three inviscid .. Case S18 (M..... 39.... Case S6 (M... Pitchingmoment coefficient. Case S16 (Moo = 2........ 48........... laminar) . 45.... a = 50... ........ = 3... 81 83 40.... a = 50... results .... inviscid) .. .... (d) lower surface of the fin on the compression side of the missile ... a = 50.. versus experiment ....5.. Case S17 (Moo = 3.. 49... a = 5 *.. the body diameter) ... 52........ inviscid) ... size and position .Figure 36. Experimental pitchingmoment coefficient.. laminar) ... Pressure contours...: (a) clippeddeltafin models................ 93 93 94 94 .. = 4...... 53... A model that was constructed with a larger fin . Case S31 (M.... inviscid) .. Pressure contours.. a = 50) .... A model that was constructed in agreement with the documented fin . Case S19 (M. laminar) ... Pressure contours....... 41....... Pressure contours. a = 5*........ Crossplane velocity components. (c) upper surface of the fin on the compression side of the missile............ 85 42......... Pressure contours............ 50. solvers ........... Case S29 (M.5... C.... a = 50): (a) upper surface of the fin on the expansion side of the missile... inviscid) .... C....... ............... Pressure contours... and EAGLE results. for clippeddeltafin model: laminar .. experiment. = 3...... Case S6 (M... xi .... = 4.... Page 79 80 80 38.. Comparison of pitchingmoment coefficient... Case S30 (M... 44. a = 50... for clippeddeltafin model: inviscid results: (a) sharpnosetip computations versus theory... laminar) ......................... 90 91 91 92 92 46.... Pressure contours........... 86 88 89 43.... = 2.. (b) sharpnosetip and bluntnosetip computations ...... Pitchingmoment coefficients from the correct fin and the larger fin Pitchingmoment coefficient. (b) lower surface of the fin on the expansion side of the missile...... a = 5*.........5. 47...
.... 69.... = 2..... ...... = 2..... for clippeddeltafin model: turbu.. Surface pressure from a sweptbluntfin/flatplate interaction and the current thinfin/missilebody interaction . Co...... 57. Cd...... Pressure contours for thickfin model (M.. = 2 and a = 50) xii ........................... 59...... a = 5°..... 70. Case S28 (M00 = 2... deltafin missile ...... lent results versus experiment . a = 50) .... Pressure near fin/body junction: thinfin versus thickfin computation 0 ) ..... Pitchingmoment coefficient....... 100 100 Pitchingmoment coefficient.. Cmo.. C.............. 124 computation .. 71. = 6... 67....... 66......... = 6 and a = 50) Streamlines in the plane of the windside fin (M. = 2 and a = 5 Pressure near fin/body junction: experiment versus thick....... Case $10 (M.. 64. a = 5'. Case SB7 (M. C..................... = 2 and a = 50) ....... Pressure contours. laminar) ..... a = 50) .... Pressure contours on the missile base...... Surface pressure coplanar with fin centerlines (M... . clippeddelta fins (a = 00) .. 102 103 104 105 65. 107 113 114 115 119 120 122 68. .. laminar) ..... Boundary for an attached shock at the leading edge of a blunt fin......... Case S20 (M...m... 72.. a = 50.. 58. 55.. for clippeddeltafin model.... inviscid) ...... Pitchingmoment coefficient..... Pressure contours. turbulent results versus laminar results: (a) sharp nosetip... for sharpnosetip HART missile with . 60......... a = 50.... 63.......... turbulent) . Boundary for incipient separation at the leading edge of a blunt fin. Pressure contours.... 74. 56...... = 2.. (M..... = 6..... comparison between deltafin missile and clipped. (b) blunt nosetip Crossplane velocity components... Pressure contours....c = 6... Case S28 (M....... Pitchingmoment coefficient.. Case S16 (M. .. comparison between deltafin missile 105 and clippeddeltafin missile .fin 122 . 73.... a = 50) Positions of the clippeddelta fin and the delta fin relative to the base Drag coefficient.. Page 95 95 97 98 99 . 62. Case S16 (M........Figure 54. Crossplane velocity components...... 61.
... altitude = 30km....... 90..... Case S28 (M_....... = 7.0...... altitude = 30km... Surface pressure.......... 152 154 Solution for highlyblunted... = 5..... 146 148 148 150 151 151 Comparison of surface pressure for modified solution procedures ........... 126 129 139 139 ........ and cold wall ... 82... ....... ..... = 0......... altitude = 30km.......... adiabatic wall) (a) pressure. ........ Case 1 and Case 7 . = 7. 89...... 92....... 80. 61 x 61 grid for hemispherecylinder .. Pressure contours........ tangentogive/cylinder (Moo = 3.. Pitthingmoment coefficient.. and cold ...........& = 5cmn..... altitude = 30km.000....... altitude = 30km........ R& = 5cm........ R. (c) Case 7..... Nondimensional pressure and density contours (M.... computed versus theoretical: M.......... Effect of nose radius on computed stagnation heating: Mo.... 5612 grid points): (a) pressure contours.. wall ................. (b) density .. R.......... Heat transfer distribution.................. 93..... 91.. ReD = 259........................ = 7........ 141 142 142 144 144 83.... Case 7 . for sharpnosetip.. (b) Case 3. Effect of source term on stability . Density contours: Moo = 5.. clippeddeltafin model: thickfin results versus thinfin results ...... and cold wall .... 79..0 where 0 is the body inclination angle given by Figure 16) .............. = lm..Figure 75.. ...... = 2 and a = 50). 88...... .. (b) thin fins (Case S28) ........ 84.. and cold wall ... a = 00. Aberration in surface pressure near the stagnation point (a = 00) . Pressure contours for thinfin model.8 (0 = 7r .. 78. Page 124 125 77.. 76.. ... Computed skin friction and heat transfer coefficients: M".. 85.. (d) Case 10 . (b) surface pressure . xiii . ..... Surface pressure versus angular position: M....... = 2 and a = 50) Crossplane velocity components (M.. Experimental/computational models and nosetip shapes ... (a) thick fins. = 5.. C..............66cm.. 87..... ...... Sensitivity of stagnation pressure to refinement of node spacing near the axisymmetry line . 81.. 86........ . R. R& = 5cm. Impact pressure (:/D = 3): (a) Case 1..
.......... Surface pressure near the nosetip of a tangentogive/cylinder in turbulent flow: M.. 160 161 162 100...... 104.... Convergence history for forebody/base combination with local time 167 ............ ramp: Mo.. 109...... ... Surface pressure from computations for an axisymmetric compression .95... 166 166 103. (b) streamlines 163 101.......33 ... R3): (a) x/D = 3. a = 00): (a) tip R3. = ..... tip P): (a) x/D = 3... (b) z/D = 5... (b) tip F3 ...... Vorticity interaction on a tangentogive/cylinder (M".33 ...95......... Velocity profiles for tangentogive/cylinder (Moo = 2. ..... (c) ... = 2....95.... Computed versus experimental base pressure coefficient ..... . a = 0°. 98. Velocity profiles for a tangentogive/cylinder in turbulent flow with Moo = 2... 107.... 95............. Velocity profiles for tangentogive/cylinder (M. tip .......... (b) z/D = 5........26..33 ........... Nonconformal grid for axisymmetric compression ramp . Surface pressure away from the nosetip of a tangentogive/cylinder in turbulent flow: M... tip F3): (a) x/D = 3.. a = 00..... Conformal grid for axisymmetric compression ramp ...... (c) tip F3 .. Typical sixsided fluid cell from a structured grid ....... a = 0°....95 .. (b) velocity profile at z/D = 3... a = 0°... (b) tip R3.....95 and a = 0*: (a) f = 3...... = 2..... Separation on flat nose tip (tip F3): (a) velocity vectors..26 .95. Surface pressure for tangentogive/cylinder (Mc....... = 2.. 108. . 173 110....... 175 175 176 178 . (b) f = 5... 105. 178 179 106.05...26.... 102........... Velocity profiles for tangentogive/cylinder (Moo = 2.05..... 158 96. tip R3): (a) vorticity contours.. a = 00): (a) Page 157 Surface pressure in nose region of a tangentogive/cylinder (M".. a = 0*) ................. (c) x/D = 6. = 2.. 2..26.95........ = 2....Figure 94..05.. (c) x/D = 6.....= 2.. (b) x/D = 5........95.....33 ....95 and a = 00 ....0.... xiv ...... 159 97. 99.. tip P. Influence of entropy correction parameter on computed base pressure (M.95 and a = 0* ..... stepping ... = 2..05....26..= 6.. (c) z/D = 6...
....95.95 and a = 2............95 and a = 2.90 ....26.90.. = 2.95 and a = 2.. (c) E = 6...26 ... 2.....95 and a = 2.. (b) E = 5...95 and a = 2.26 ..95 and a = 2.. and ..... = 2....05..90: (a) M = 3..... 184 ... Surface pressure near the nosetip of a tangentogive/cylinder in turbulent flow: M.Figure 111.90.... = 2. and E = 3..90 .. ..... Windside velocity profiles for a tangentogive/cylinder in turbulent flow with Mo. .. a = 2..26. Page 181 181 112......... 116. = 2. = 2..90 .....05... Leeside (0 = 0*) velocity profiles for azimuthal grid refinement: Mo = ..33 183 115... Windside . a = 2. = 2. Comparison of the azimuthal variation of surface pressure using coarse and refined spacing near the body: M.33 182 114. M.. 184 117........ (c) M = 6.... = (4 xv ...... 113.90: (a) f = 3.90 . 85 186 1800) velocity profiles for azimuthal grid refinement: .. Leeside velocity profiles for a tangentogive/cylinder in turbulent flow with M.= 3... Surface pressure away from the nosetip of a tangentogive/cylinder in ..... Comparison of the azimuthal variation of surface pressure for azimuthal grid refinement: Mo = 2.......... 118..... turbulent flow: M. (b) M = 5....95..............
. 8...... Pitchingmoment coefficient........ Summary of numerical experiments on the bluntnosetip model used for grid sensitivity analysis ............ Cm..... Pitchingmoment coefficient... Pitchingmoment coefficient... . Pitchingmoment coefficient.... Page 51 53 55 4.... 2........ 7.. from current computations for sharpnosetip model with clippeddelta fins (Cases S1S34) ........... sharpnosetip model with clippeddelta fins ........ . from freeflight tests of the grooved... 3...... Experimental drag and pitchingmoment coefficients for bluntnosetip model with clippeddelta fins ..... 71 71 73 75 76 83 83 14.... Summary of numerical experiments for sharpnosetip model...... xvi ..................List of Tables Table 1............ .... 13. Cn.......... Experimental drag and pitchingmoment coefficients for sharpnosetip model with clippeddelta fins (a = 00) .................... Error between theoretical trajectory and experimental trajectory (from reevaluation of freeflight data) .. 10............ .......... 5......... 63 64 64 Computed drag components for sharpnosetip model (a = 00) ............................ ....... Computed drag components for bluntnosetip model (a = 0*) .... 12..... 56 ... Cm............... Summary of numerical experiments for sharpnosetip model...... from ZEUS and EAGLE results for sharpnosetip model with delta fins . 6... ........ 11....... without base region ..................... C... without base region ...... with base region ...... from current computations for bluntnosetip model with clippeddelta fins (Cases B1B31) ....... 9........... Drag and pitchingmoment coefficients from EAGLE results for sharpnosetip model with clippeddelta fins . ......... Summary of numerical experiments for bluntnosetip model................ ....
Summary of comparison between computation and theory or experiment 131 Summary of numerical experiments for hemispherecylinder ................... 20...Table 15.. ... ... 22..... C....0....... Comparison of computer code performance for 3D solvers: full NavierStokes and turbulent flow. 24... and center of pressure. Page 84 84 121 131 17... Comparison of data processing rates for AFITENSAXI Version 2. and center of pressure...... Normalforce coefficient. C.. from current computations for bluntnosetip model with clippeddelta fins . 16.. .... 0 ...... Normalforce coefficient...... Resource utilization .............. 23........ from current computations for sharpnosetip model with clippeddelta fins ...... ZqP. ........ Comparison of computer code performance for axisymmetric solvers.. 19. 150 187 191 191 192 xvii ...... 18.... Summary of sensitivity evaluations ... zX..... . Drag components from thickfin computations (a = 00) .... 21...
Cd = near wall region length parameter = = = axial force coefficient total drag coefficient base drag coefficient friction drag coefficient pressure drag coefficient Cd6 Cd. LC fin. k L L.List of Symbols Symbol A+ C. (per unit area per unit time) limiter for the Ith wave field flux Jacobian Jacobian of the transformation (2D) thermal conductivity body length operators for strangtype fractionalstep method unit normals in the Cartesian coordinate directions turbulence model pressure gradient function pressure pressure on the fin lower surface pressure along the fin's leading edge xviii g9 J . D et Ei+½j F = wave drag coefficient = = = = = = = lifliftt coefficient pitchingmoment coefficient () normal force coefficient body diameter total energy per unit volume numerical flux at the (i + j. k interface = = = = = = = flux vector of a conserved quantity.. Lf.. fix N p PA Pie = = = = = . CdP = = C4 C.. C. U.• C. fn.
.PU Ph = = pressure on the fin upper surface P+ Pr Prt q. y. qy. r. q. 9 X. r RN RB = = impact pressure (total pressure behind a normal shock wave) Clauserlike pressure gradient function (inner formulation) Prandtl number = turbulent Prandtl number = = = = heat fluxes in Cartesian coordinate directions radial coordinate nosetip radius body radius eigenvector due to diagonilation of flux Jacobian Reynolds number per unit length Reynolds number based on length A Re RCL s S(X) S1 S. v. w U V XCO XCP X. z i y+ Z = = = conserved quantity (per unit volume) = cell volume = = center of gravity center of pressure = cylindrical coordinates = Cartesian coordinates = = distance from nosetip in axial direction nondimensioonal distance normal to surface 1/2 = turbulence model compressibility function xix . SB = = = = distance from nosetip along surface = crosssectional area at x = fin planform area = = wetted area base area temperature Cartesian velocity components T u.
Tzz TZz.. rT. x. k interface /3 = = = = = fl" 6 At shockwave angle Clauserlike pressure gradient function (outer formulation) boundarylayer thickness time step in numerical integration coefficient in entropy function ratio of specific heats eigenvalue due to diagonalization of flux Jacobian . 7 = density TI~.y A = = = entropy corrected eigenvalue A A A•t V = = = = fin leadingedge sweep angle first coefficient of viscosity (molecular viscosity) eddy viscosity Courant number (L azimuthal angle (also potential perturbation variable) upwind function S= = p Tzz.a = = angle of attack vector of characteristic variables at the (i + j. = viscous stresses xx .
Ultimately. laminar. this results in lower fin effectiveness and reduced static stability. The formulation models turbulence with the zeroequation. Turbulence and the blockage phenomenon cause bleeding around the fin leading edges. In addition. The equations are solved using a finitevolume methodology. In the first part of the study. and turbulent conditions and Mach numbers from 2 to 6. BaldwinLomax turbulence model. algorithm is implemented and employed to solve the NavierStokes equations. many unresolved issues from experimental testing of the HART missile are addressed. xxi . A comparison is made between thinfin results and thickfin results to assess the impact on missile stability. The present results also suggest that the clippeddeltafin configuration is stable beyond Mach 7. In addition. In the second part of the study. The aerodynamic and staticstability characteristics are investigated to determine if conventional supersonic missile configurations can be flown at Mach numbers higher than 5. the effects of fin thickness and cross section are explored. The effects of nosetip blunting and boundarylayer condition are demonstrated. The structure of the flow near the fins is significantly affected by the turbulent transport of momentum in regions of blocked cross flow. numerical experiments are performed using an infinitely thinfin approximation. The current predictions indicate that the pitchingmoment coefficient decreases with increasing Mach number much less than previous numerical computations. secondorderaccurate. the strength and extent of the flow structures that develop in the blocked regions appear to be enhanced by fin thickness. fluxdifferencesplitting.AFIT/DS/AA/943 Abstract The flow about the complete Hypersonic Applied Research Technology (HART) missile is simulated for inviscid. An explicit. The aerodynamic characteristics of the HART missile are predicted at Mach numbers beyond the experimental freeflight testing capabilities. accounting for pressuregradient and compressibility effects.
85] has looked extensively at the aerodynamics of spinning shells (with and without fins). The literature is replete with studies of turbulent boundary layers in supersonic flow. and magnus characteristics [44. Additionally. 82. To realize the tactical advantages associated with increasing missile speeds. 83]. 13. Finally. with respect to modeling turbulent boundarylayer development and stability performance at very low angle of attack are rare. very high finenessratio finned bodies. Weinacht. 1. an overview of the dissertation is made to lay a foundation for the chapters which follow. In fact.1 Background on Missile Research There have been many calculations for both axisymmetric and finned projectiles. no information could be found outside thesir Force program to be discussed shortly. boundarylayer development. Investigations of unspinning. 84] are just two examples.AN AERODYNAMIC AND STATICSTABILITY ANALYSIS OF THE HYPERSONIC APPLIED RESEARCH TECHNOLOGY (HART) MISSILE 1. The scope of the present study relies on previous work in this area. a description of the purpose for the present study. spinning and finstabilized projectiles in supersonic flow. and an overview of the scope of the work. Several studies have been made to assess the effect of nose bluntness on surface pressure. the Air Force Armament Laboratory (now the Wright Laboratory Armament Directorate) initiated an inhouse program to investigate potential methods to extend 1 . It begins with some background into Air Force highspeedmissile research as motivation for the current investigation. Sturek [81. et al. Introduction This chapter presents a review of previous research. some background is provided on the fin/body interaction problem. [98] have conducted PNS computations for much lower finenessratio. [30. His work has concentrated on the influence of shell geometry and yawing on the magnus effect at supersonic speeds.
848 Z. y" 21 82 L0. 24. The goals of this phase were to predict. has concentrated on varying geometry parameters like nosetip shape and fin shape. 2 . cylindrical body.(Horizntal (a = 0.flight regimes for future airtoair missiles. tailcontrolled missile. which began in 1987. The Aeroballistics Section of the Wright Laboratory Armament Directorate developed a configuration that would most dramatically exhibit the staticstability limitations of current missile designs. and one fourfin group located near the extreme aft end of the projectile (shown in Figure 1 with a sharp nosetip and delta fins). The variations to the baseline model of Figure 1 will be discussed in the following sections together with the experimental and numerical results.0 Figure 1.) Symmetry Planes Plane) 1. and was named the Hypersonic Applied Research Technology (HART) missile.. A corollary goal was to begin identifying key missile geometry parameters that will effectively increase the stability envelope.251 y U.0 2. It was a generic high finenessratio (length to diameter). x __i _1 0. The first phase of this program involved a detailed investigation and documentation of the flow physics affecting the static stability of missiles. Hypersonic Applied Research Technology (HART) baseline model Experimental testing. substantiate. The model consisted of an ogive nose. and understand the staticstability characteristics of a generic missile configuration. and continues to date. This configuration was intended to provide the baseline characteristics from low supersonic to hypersonic velocities.
the nosetip shape was modified with blunting.2 [22]. There are 131 locations available as instrumentation sites. The Aeroballistic Section of the Wright Laboratory Armament Directorate made use of several prediction techniques.1. and numerical. The ARF is an enclosed concrete structure used to examine the exterior ballistics of various munitions while in unrestrained flight [45]. neutral stability was predicted and confirmed to occur at about Mo. These are subsequently used to determine the spatial position and angular orientations of the projectile at each of the shadowgraph stations. empirical. The code was developed by the Naval Surface Warfare Center and called 3 . First. and are identical to the configurations analyzed in this study.1. a model with smaller delta fins was also investigated to confirm that the loss in static stability can create an unstable situation at high Mach numbers. = 2.1. PRODAS is an interactive program which allows the engineer to create a projectile. In 1989 and 1990. Two numerical methods were utilized. while maintaining the fin span of the small deltafin model. which is compared to similar configurations.1 Experimental Testing. 4. an inverteddeltafin model was constructed with the same area and span constraints described above. At the time the baseline model was tested. spacemarching Euler code was used. Eglin AFB. The empirical predictions were made with the Projectile Design Analysis System (PRODAS) [36]. including analytical. and used to obtain orthogonal photographs of the model's shadow during flight (called shadowgrams or shadowgraphs). In addition. It uses the Newtonian Flow Method [21 and the Second Order Shock Expansion Method [2] to calculate flow parameters. FL. the baseline model showed a decreasing but still stable (negative) pitching moment up to the highest Mach numbers tested (M. three additional models were tested [93].2 Analytical and Numerical Testing. The sharp and blunt clippeddeltafin models are shown in Figure 2. an inviscid. In contrast. The analytical tool is entitled "Missile DATCOM" (94. Freeflight tests were conducted at the USAF Aeroballistic Research Facility (ARF). A clippeddeltafin configuration was fabricated with the same fin area as the baseline model. to corroborate experimental findings. In addition to fin modifications. The instrumented section of the ARF is 207 meters long. . It was expected that the static stability would be improved by moving more of the fin outside the boundary layer [93]. The program then interpolates an extensive data base to provide aerodynamic estimates. 1.5) [22]. For the small fins. 951.
82 Li Figure 2. a multiblock. 0. steadystate Euler solver was used. then solves the steadystate Euler equations. Unless otherwise stated.5 for a discussion of the construction of a computational grid): * Roe flux differencing * Characteristic surface boundary conditions * 40 points from the outer domain to the body * Grid spacing clustered near the body with a minimum spacing of 0.004D e 19 points for the azimuthal halfplane e Equal azimuthal angular spacing of 100 * Grid conformed to the fin leading edge through block structure e Fin was treated as infinitely thin Solutions were computed for the baseline HART model (delta fins). 88.40. 87.. pertaining to the EAGLE solver. This solver actually comprises a family of programs called the Eglin Arbitrary Geometry Implicit Euler (EAGLE) Code [50. 60].25~.1 X .0 2. The EAGLE solver constructs a multiblock grid. and the HART model with two other fin configurations (clippeddelta fins and inverteddelta 4 . implicit. HART models with modified nosetip and clipped delta fins the Zonal Euler Solver (ZEUS) [97]. the following options. were selected for all data presented in this document [68. Finally.y 24.0 I2. 69] (see Section 3.
54. To date. experimental. is the wetted area..58 b (2) where S.50 [68]..fins). the numerically predicted drag did not correlate well with the experimental drag.e 4. 1.3 Summary of Previous Results. with increasing Mo. The following were used for base drag and skinfriction drag [68]: Cd6 Cd. The resulting total drag appears in [68] as well as [93]. The numerical drag prediction (EAGLE) in Figure 3 is empirically corrected for base drag and skinfriction drag. In particular. The EAGLE solver computed a greater decrease in C. Finally.5). this was not resolved. The other components were calculated empirically. For each of these configurations a solution was calculated at seven different Mach numbers for a = 0 and a = 0. The results from prior testing on the baseline model are presented to highlight a couple of these problems. . Second. = K [log(ReL)]2. ZEUS and EAGLE predictions of the pitchingmoment coefficient differed significantly above Mach 5.. The experimental data is depicted with open symbols in both figures. freeflight testing showed an unusual trend in the pitching moment.2 [93].0) [93]. The drag and moment results are tabulated with the results from the present study in Chapter 4. and Sb ik the base area. empirical. This was not expected and not resolved by the researchers of [93].. The poor correlation between the experimental and 5 . particularly above Mach 3. Again.1. and computational analyses of the HART missile have had varying degrees of agreement. The pitching moment is directly output from the EAGLE code. the observed drag exhibited larger than expected scatter over the Mach number range tested (1. These are the central questions to be addressed in this document. the clippeddeltafin and inverteddeltafin missiles appeared to become more stable in the highsupersonic regime (M. differences between results for the modified geometry and the baseline model tested in 1987 were not addressed in [93]. four unexplained discrepancies materialized out of the previous combined experimental/computational work.. and much lower stability margins. Since EAGLE is an inviscid solver. only the forebody wave drag can be obtained. Further. Third. First. The drag coefficients and pitchingmoment coefficients obtained from experiment are plotted with the numerical results in Figure 3 and Figure 4.
quantify the effects due to nosetip shape. They found that the flow remained attached for sweep angles exceeding 68 degrees.71 or Moo = 5. nobody has investigated turbulent. as the fin sweep angle increases. some questions remain as to whether separation occurs for severely swept fins at higher Mach numbers. To the best of the author's knowledge.45. Third. base region. viscosity. develop an efficient computational tool to simulate accurately the myriad of flow structures associated with highspeedmissile flight. In Figure 4. Recently. but the disagreement between the numerical predictions becomes large at higher Mach numbers. Price and Stallings [64] reported the effects of sweep as long ago as 1967. while substantiating the aerodynamic traits of the HART missile. however. 1. flatfaced fin.1. investigate the unexpected variations in the aerodynamic coefficients obtained from experimental testing (explained in Section 1. 6 . First. the flow structure which should be predicted for this fin geometry is not entirely clear.98. there are very complicated flow structures associated with a blunt fin attached to a body in supersonic flow. and can involve the formation of strong horseshoe vortices [80]. Next. McMaster and Shang [55] used a McCormack method with the BaldwinLomax turbulence model to simulate a blunt fin on a flat plate at M. 1. the experimental and numerical (EAGLE and ZEUS) pitchingmoment coefficients agree at the lower Mach numbers. He confirms the benefit of sweep in reducing the level and extent of the disturbed pressure field.3 Purpose for Present Research There are three main purposes for the present study. Price and Stallings [641 and Winkleman [99] experimentally found no disturbance ahead of a 75 degree swept fin at either M. = 3. Hussain [41] studied blunted fins at sweep angles up to 75 degrees.3). The fin's bow shock interacts with a growing boundary layer on the body. The interaction weakens dramatically. = 2.numerical drag is evident as Mach number increases. supersonic flow past a swept. Thus. Although Hussain found clear evidence (oil flow pictures) that separation occurs up to the highest sweep tested at Mo. fin thickness. = 2.2 Background on Fin/Body Interaction Problem In general. and turbulence.
. 5 6 7 8 Figure 4.70  . C~o~so 0.° . 4 M. 10 1 I I I I I 2 3 4 M.I .. I 2 3 ..0. 5 6 7 8 Figure 3. 30 20 10 .. N. ZEUS C50 40 'S.utral Stability" .40 0. Axialforce coefficient (drag) for HART baseline model 110 100 90  cExperimu 70 EAGLE 8.90 ' 0.. AGE . .8t0  0aa U 0.30 1 ... Pitchingmoment coefficient for HART baseline model 7 .
The scope would be undesirably impacted because the run time for each 8 . requiring accommodations for larger variations in the trajectory and potentially unstable or very dynamic tests. For example. Clearly. and provides the means to study the flow physics and missile performance of future tests with greater detail than obtainable from freeflight testing. supersonic flow and shocks. Computationally. To overcome the staggering memory implications. we can explore configurations and conditions beyond the practical limits of ARF testing. For configurations which have length scales that differ by several orders of magnitude. Florida. flight model failure currently occurs during launch at Mach numbers above 4. Since highspeedmissile research entails accurately predicting each flow structure mentioned above. When the simulation involves turbulent boundary layers and wakes. Also. and shock/boundarylayer interactions.1. Therefore. As discussed in Section 1. but a different approach was attempted because this would have drastically reduced the scope of the investigation.5 [93]. the results and analyses from this research will be used by missile designers to predict weapon performance characteristics in the hypersonic flight regime. a tradeoff in algorithm performance for memory efficiency could have been pursued.4 Overview of Dissertation Simulation of a complete configuration is demanding (minimally requiring O[100K] grid points). 1. In combination with the experimental data. future configurations may fly at higher angles of attack and possess different centerofgravity locations. the grids here could easily be millions of points.The development of an accurate and efficient computational model is essential for the above investigations. accurate results require sophisticated modeling techniques and many more grid points: O[250K] points. and further the understanding of the fluiddynamic mechanism affecting missile stability at higher Mach numbers. freeflight experiments are conducted in the Aeroballistic Research Facility (ARF) at Eglin AFB. The intent is to substantiate the aerodynamic traits of the HART missile. and the ratio of missile length to fin thickness for the HART missile is about 500:1. the present research is driven by the results of the prior testing.1. this work will address all of the previous testing in the course of the investigation. O[1000KI grid points are usually necessary. and is aligned with the previously mentioned Air Force research program.
This can be confirmed quite adequately on a reduced scale. In regards to computational analysis. = 2 . Lowering the grid requirement is achieved with two assumptions or approximations. Several serniempirical corrections are proposed to account for the lack of fin cross section.numerical test increases with this type of tradeoff. the current computations are qualitatively compared to the computational results of the two inviscid solvers used on the baseline model." The assumption relies on the premise that fin thickness is secondorder to fin impermeability. an infinitely thinfin approximation was made. In Part II. to a.4. The alternate approach is to reduce the grid requirement. 1. Stability is investigated over the Mach range with the emphasis on the impact of viscosity. Although most grid refinement issues are addressed prior to testing these shapes (see Appendices A and B). The fins are treated as having no thickness. but. and using the thinfin approximation. and turbulent conditions for the sharpnosetip model shown in Figure 2. In Part I. the effect of turbulence. numerical experiments are undertaken to determine the aerodynamic coefficients for the two geometric variations of the HART missile. and sensitivity of C. Parametric Mach number studies (Ma. laminar. the analysis is extended to higher Mach numbers than obtained in freeflight tests. Also. The analysis is formulated to address discrepancies that developed out of previous combined experimental/computational work on these configurations (Section 1. After this sensitivity study.6) are conducted for a = 00 and a = 5° under inviscid.. Computations are made for the sharpnosetip and bluntnosetip. owing to the projected grid refinement.1 Scope of Thinfin Investigation (PartI). The effect of the fins on base drag is also explored. 9 . For high Mach numbers.1. a limited grid refinement and sensitivity study is made with the thin fins to evaluate the variation of drag and pitching moment with azimuthal grid spacing. the complete configuration is not reinvestigated. this is not an example of Keller's principle: "Something for nothing is worthless. and the influence of the base flow. Recall that these results differed significantly above Mach 4.0 (Figure 4). clippeddeltafin HART missiles (Figure 2) using the infinitely thinfin approximation. The drag is broken down by component to facilitate various comparisons to the trends with Mach number predicted by theory and other results. the actual thickness of the fin is modeled. and so they are merely boundary conditions within the computational domain.3). while maintaining high algorithm performance.
and identify key missile geometry parameters affecting the stability envelope. The flow structure is directly compared to the structure from the thinfin computations. In addition. 10 . and thereby validity.2 Scope of Thickfin Investigation (PartII). and is explored with a parametric Mach number study. Next. the impact of fin thickness on static stability is addressed. computations are made using the modified geometry to further document the experimental discrepancy. The aerodynamics of the clippeddeltafin HART models are contrasted with the baseline model in an attempt to address the goals of Air Force hypersonic missile research. of the thinfin approximation made in Part I is addressed in this part of the research.4. the structure of the flow is examined to determine if fin thickness significantly affects the interaction at the fin/body junction. Finally. The accuracy. and its impact assessed. This study addresses several questions relevant to the blunt models: Is the dependence of the pitching moment on angle of attack affected by nosetip geometry? Does the lower total pressure associated with the normal portion of the bow shock wave affect pressure near the fins? Does delayed transition associated with the high entropy layer on blunted bodies impact the HART missile's pitching characteristics? A discrepancy in the sharpnosetip experimental models is documented. fin drag is compared to results from the thinfin computations.The aerodynamics of the bluntnosetip model are also determined with a parametric Mach number study. 1. the goals are to substantiate the stability characteristics of high finenessratio missiles. A boundary for incipient separation at the fin leading edge/body junction was suggested by Stollery [80]. First. As previously stated.
Part I: Investigation Using "Infinitely Thin Fin" Approximation 11 .
2. and the pitching moment on a finned body. the relation between the extent of a shock/boundarylayer interaction and leadingedge bluntness.2. The fluid equations are presented in full. Figure 6 shows a relationship between the shockwave angle and a series of delta wings. and incidence angle is highlighted.. Next.1. and the intersection line is swept. In particular. and turbulence is accounted for with the Boussinesq [101] eddy viscosity concept. Shock Wave Shape. threedimensional form. the only shock wave of interest is that formed by the shock generator (fin). on missile body). and the only boundary layer of interest is the one affected by the glancing impingement (i. therefore. Figure 6 is taken from [801 (with permission) to illustrate the shockwave shape expected. The generated shock and the boundary layer of interest are shown in Figure 5.1 . the fin is swept at an angle A. In general. In the simplest interactions. the deflection angle depends on the crosssectional shape and the incidence angle. sweep angle.e. inviscid drag on a slender body. the governing equations are presented with a brief discussion of boundary conditions and turbulence. Finally. inviscid theories for drag and pitching moment are examined. the flow structure near a fin/body junction is discussed.1 Flow Structure Near Fin Glancing interactions include situations where the shock wave generated by one body cuts across the boundary layer growing over another. Governing Equations and Supporting Theory The equations of fluid mechanics that are fundamental to the present research are outlined in this chapter along with important principles for the fin/body problem. The deflection 12 2. The introduction of fins onto a forebody is a departure from the validation models (axisymmetric bodies). Figure 5 shows the geometry of a swept thin fin. Eddy viscosity is invoked to achieve closure of the Reynolds equations for turbulent flow. First. and inclined to the freestream at an angle a. linear theory is used to provide a bridge between the simpler validation computations and HART missile computations. Stollery [80] gives an excellent review of glancing shock/boundarylayer interactions. The delta wings are wedge shaped with a wedge deflection angle indicated by a in Figure 6.
however. sweep (A). and a very high 13 . the associated vortices lose strength or do not develop at all.1. is exactly equal to the incidence angle. separated. In addition there is a very low incidence angle (a < 50). and V is the core of a vortical structure. It is particularly interesting to note that the shockwave angle should be very close to the Mach angle (p in Figure 6) for the typical HART missile fin shape (A . [80]. therefore. A is an attachment line. Streamsurface and vortex skeleton patterns are shown for attached. 2. and reducing the fin's inclination to the oncoming flow. increasing the sweep angle.shock r Figure 5. and orientation (a). S is a separation line.2 Shock/Boundarylayer Interaction. Swept thin fin geometry angle for a thin fin. Figure 7 is also taken from [80] (with permission) to establish a basis for discussing the interactions.70*) at low angle of attack. This is usually taken to be the angle of attack. and secondary separation flows. In this study. The strength of the shock/boundarylayer interaction depends on the fin's leading edge radius of curvature. reduces the extent of the interaction: reducing the leadingedge bluntness. In turn. In the vortex skeleton representations. Each of the following. but in this investigation it is somewhat less. since the fins are placed at 45 degrees relative to the horizontal (Figure 1). the thinfin assumption eliminates the radius of curvature at the leading edge.
42 * + j 404 Wspi./ia / " .'/ :.9 N " // .• !. Figure 6. Shockwave angles measure from the centerline of a delta wing (from 180] with permision) 214 152 .2 I I.
sweep angle (A ; 700). Thus, the flow pattern should agree most closely with the attached representation of Figure 7. 2.2 Inviscid Theory for Drag and Pitching Moment Inviscid theory provides a suitable comparison for computational results. By examining simple theoretical expressions for wave drag, an appropriate baseline can be established by which inviscid and viscous computations can be interpreted. In addition, by using theoretical aerodynamics for the fins, the pitchingmoment trends with Mach number and angle of attack are formulated. These qualitative baseline trends are used to interpret and understand the computational results. £2.2.1 Theoretical Forebody Wave Drag. The pressure drag acting on a body in supersonic flow can be thought of as composed of three parts: the wave drag, the vortex drag, and the base drag; the drag coefficients are C 4 , Cd., and Cd., respectively. The vortex drag arise,_ from the momentum carried away by vortices trailing from a lifting body [4]. No vortices exist for the HART missile configuration at the angles of attack examined herein. A theoretical solution to the base drag is not available, so empirical relations are often used of the form
Cdb =
K
(3)
where K is an experimentally measured function of the freestream Mach number, M., [56]. The wave drag results from the momentum carried away by the pressure waves set up by a body as it travels faster than the speed of sound [4]. Slenderbody theory [4] provides an expression for wave drag in terms of a body's profile of crosssectional area, S(x), and the solution to a perturbation potential equation. By virtue of the smallness of the disturbance (slender body), a perturbation is introduced into the full potential equation (FPE). A linearized equation is derived to govern the perturbation potential (see [4]). The solution to this linearized equation provides 4 and g(x) in the following expression for wave drag:
D  Db
DDU
L
1 
1da
= jo
(x)g(x)dx

1S(L)g(L)  j
(4)
In (4), L is the length of the body, and Cb represents a contour around the base. 15
Introducing the solution to the perturbation potential equation for an axisymmetric body, and recognizing that for a body which ends in a cylindrical portion parallel to the freestream, 8•k/dn and S vanish, yields D  DA _ 2TV
1
PooU.2
L §()d
oS(x)ln(z XI)dxl,
(5)
or
Cdb =
r
I S(z)dx
S(xi,)ln(

xi)dxi.
(6)
The most striking aspect about (6) is the absence of Mach number. This agrees fairly well with experimentally determined wave drag on very thin bodies of revolution [56]; wave drag on these bodies shows very mild Mach number dependence. The HART missile has a 2.25 caliber tangentogive nose (Figure 1). Drew and Jenn [15] numerically evaluated (6) using the crosssectional area distribution for this nose shape. They found Cd, = 0.232. This value is somewhat high compared to experimentally determined wave drag [56], ranging from 10% high at M.  1.5 to 50% high when Moo > 6. The overprediction is a result of the relative bluntness of the nose. Slenderbody theory performs better for fineness ratios exceeding 3. The error associated with the slenderness assumption can be overcome by using a higherorder analysis, or by an empirical correction, and the Mach number dependence can be interjected by using the frontal projection area of a modified cross section. The new cross section is created by intersecting the body with a plane at the associated Mach angle. This is Whitcomb's area rule [100]. Both Mach number and fineness ratio were accounted for by Drew and Jenn [15]. At M". = 2, they calculated the wave drag at approximately 0.166. (This is in reasonable agreement with the current calculations shown in Chapter 4.) 2.2.2 Pitching Moment From Linear Theory. Finstabilized configurations are known to exhibit decreasing static stability with increasing Mach number. The reason can be traced to a loss in the fin effectiveness associated with compressibility. This basic principle is clear from thinairfoil theory. For a flat plate, at angle of attack, in supersonic flow, the lift coefficient determined from thinairfoil theory is [4]
S(74a)
16
From (7), at a fixed angle of attack, as Mach number increases, the lift created by a flat plate decreases. Since the fins are treated as swept flat plates, the lift from the fins also decreases. Although boundary layers exist, and shock interactions affect the solution, this underlying principle dominates. In contrast to thinairfoil theory, slenderbody theory predicts that the lift from the missile body is independent of Mach number [4]. Ci = 2a. (8)
Thinairfoil theory can be modified to account for the azimuthal locations of the fins, and combined with slenderbody theory to predict the lifting and pitching characteristics of a missile. Basically the more or less constant lift associated with the unfinned missile surface produces a nearly constant destabilizing moment. From (7), as Mach number rises, lift from the fins decreases. The lift from the fins is modified to account for the nonhorizontal placement of the fins on the body (Ofin # 900), and normalized to the missile base crosssectional area, Sb. The ratio of the HART model's clippeddeltafin area, Sf, to base area is 1.2223. Because the fins are aft of the center of gravity, less lift means the restoring moment decreases. Thus, the static margin or pitchingmoment coefficient approaches neutral stability. The functionality of the pitchingmoment relationship is obtained by using (7) and (8) with appropriate moment arms. The moment arm for the lift created by the body is the distance between the center of pressure for the body alone, Xp,,, and the center of gravity for the total configuration (body with fins), x... The moment arm for the lift generated by the fins is the distance between this same center of gravity and the center of pressure for the fins alone, xzCP. becomes
C = c
C2
The pitchingmoment coefficient (9)
where cl and c2 are
cl= 2 (X' zCP 1)
and
C 'b2 D Xc 9 ) (4S 1 ) 2 = 4 ( Xq'
(cos
Ofin1).
(10)
The constants cl and c2 are calculated to correspond to the center of gravity at 43.5% of length. The center of pressure for a fin is taken to be its centroid. Therefore, 17
XM  Xcg = 11.906D.
The center of pressure for the body is calculated using the
experiments of Butler, et al. [9] and Dolling and Gray [13] on tangentogive cylinders. Butler, et al. and Dolling and Gray tested much shorter bodies than the HART missile body, but all results showed that for small angles of attack, surface pressure had equilibrated within six diameters downstream of the ogive section. Therefore, pressure was assumed constant over the remaining length of the HART missile body,
and xc,  XCg = 8.440D. Figure 8 is a graph of (9).
"90
80 70
50
C
40 30 20r
uEquion (9)
10
0
1lo
h
2
Ncutral Stability
I
I
I
I
3
4
M.
5
6
7
8
Figure 8. Pitching moment using thinairfoil and slenderbody theories
2.3
Governing Equations and Boundary Conditions
Equations are presented throughout in nondimensional form. The velocity and density scales are the freestream velocity and density, U,, and p,,, respectively; pressure is nondimensionalized by p..U.. The length scale is the missile body diameter, D; and time is nondimensionalized by the aerodynamic time scale, D/U,,. 18
. The integral form of the NavierStokes equations is written as / U dV + hi. Only modest assumptions are involved in (11). + wryz Pw  i ) puw . and the surface of that region is S. F dS = 0. conserved quantities cannot be produced internal to S. Also. First. Stoke's hypothesis is used for the second coefficient of viscosity to yield the following flux definitions in each Cartesian direction [1]: pu pu 2 + p  Re1T F E= puvRe'm.'1) Pv puv . + V 2 + 19 . ii F G= pvw. The volume of the region over which (11) is applied is V..Rer1 2 (et + p)w  Re`1 (Ur.RerTy pw 2 + p . umed..Relr. (11) where U is a conserved variable (per unit volume). Next. f = F= (et + pv2 + p 1 ReTrY (12) pvw . all variables are assumed continuous in time.. puw . + VT.Re1.xy + W'rz  (. Firlly. Newton's assumption that a linear relationship exists between fluid stress and strain is a6.Re` (uT v+r +1 . (et + p)u  Re (uT. as with a chemical reaction for example.Re'rx_.The appropriate governing equations for a perfect gas are the NavierStokes equations. a continuum is required. and F is the flux of U (per unit area per unit time).Re"r• p)v .
which is taken to be the standard sealevel temperature. For laminar viscous computations. The equation of state for a perfect gas defines pressure (nondimensional): p = (f1)e = (7 .4K)/Too. Viscosity and thermal conductivity are nondimensionalized by their freestream values. Re. is equivalent to p/Pr. The freestream values are referenced to the freestream temperature. = T + ~u a eaw w '9 .The Reynolds number. those terms preceded by Re1 are set to zero. = (110. S=kaT q = kTZ 9• q=kad and q q..1. p is determined using Sutherland's formula [1]: 2T1 (i •. represents the ratio of convective to dissipative forces.) C =+ . (16) where c. . Pr = 0. k (14) In nondimensional form the coefficient of thermal conductivity.) Ip P * (15) For inviscid calculations.=x. To. The subscript 1 denotes that (16) represents the molecular 20 . k. q are defined as a. The components of the heatflux vector. a is the the first coefficient of viscosity.71. The viscous stresses are defined as follows [1]: E 4av = 2 (au +8w\1 x au 3 o~ 3 Oz)JI (13) 4aw 3 T..1)(e. ( iv In (13).
and the model was inclined within one of the planes of symmetry. Then. axisymmetry is applied to surfaces 7 and 8 in Figure 9. Also. freestream conditions are applied. and fin surfaces. base (surface 4 in Figure 9). Consequently. "P%JAI + Pt. Equations (11)(14) do not directly govern a turbulent flow. at the inflow (surface 1 in Figure 9). calculations were made around only half the circumference of the missile. Due to the supersonic condition at both the inflow and outflow boundaries. However. and at the outflow (surface 2 in Figure 9). Since the HART missile is bisymmetric. if the Reynoldsaveraged NavierStokes equations are cast in nondimensional form. then (11)(14) still apply with minor alterations [101].9 [101]. Prt. the governing equations are used in nondimensional form. pt. is invoked to achieve closure. More details are provided for these boundary conditions in Section 3. on the body. The turbulent Prandtl number. Symmetry is then enforced at the azimuthal boundaries (surfaces 5 and 6 in Figure 9). pt is calculated only in the turbulent regions. An empirical model is used to calculate pt in the wake. Recall.6. The turbulence models and the location of the laminar to turbulent switch are described in Section 3. A location is selected along the missile length to separate laminar and turbulent portions of the boundary layer. For the body (surface 3 in Figure 9). the associated boundary conditions are relatively straightforward.viscosity for laminar flow. zero normal pressure gradient and an adiabatic wall are enforced. base. For a = 00. k = Arr + Prt' and pi is still calculated with (16).4. and fin surfaces. for air is assumed to be is 0. impermeability and no slip (viscous only) are enforced. and a fully turbulent condition is assumed for the wake region. The computational domain is illustrated in Figure 9 where each boundary is given a number. Otherwise extrapolations give necessary data at those boundaries. extrapolation is employed. and eddy viscosity. The BaldwinLomax algebraic model [5] is used to compute pt on the turbulent portion of the forebody because of its relatively low computational cost and ease of application. 21 .
some secondorder coefficients were determined (e. Now the drag coefficient becomes: Cd = Cd. 93]. derivative coefficients are calculated numerically. C.g.4 Aerodynamic Coefficients Both static and dynamic firstorder force and moment coefficients were obtained from freeflight experiments of the HART missile configuration 122..Figure 9. + CdQ + Cd. 22 . + Cd...1 Drag Coefficient.. 2. C4. In terms of a drag coefficient this takes the following form: Cd = Cd. Cd.o). and the change in pitchingmoment coefficient with Mach number. The experimental and computational methodologies are explained in the following two sections. The forces acting on the missile can be separated into pressure and viscous contributions. Computational domain showing boundary conditions 2. Additionally. the change in zeroyawdrag coefficient with Mach number..4. The present computational study provides only axial and normal force coefficients and the firstorder pitchingmoment coefficient. The pressure drag can be further separated into a wave drag and a base drag.
Fin base drag is determined by correlating the local Mach number and pressure near the fin trailing edge with the pressure for 23 .) Unfortunately. (See Chapter 3 for a description of the present methodology and grid limitations. 2. This portion of the drag is not directly computable due to the thinfin approximation. To make any quantitative comparison with experimental data meaningful.2 Fin base drag. + Cd. and base drag contribution.4. but due to insufficient grid resolution and lack of fin cross section. viscous. fin drag must be accounted for.where the fins and the missile body each have a wave. Impact pressure is the total pressure behind a normal shock wave sitting in front of an imaginary numerical probe inserted into the flow. 4. The following steps are used with the current calculations to obtain a value for fin wave drag. pie. Previous EAGLE computations were also made with the thinfin assumption. 2. pie = 0. Obtain the pressure along the fin's leading edge. Use (18) with pie and the actual thickness of the fin to determine fin wave drag. For A z 700.2P/. + Cd + Cd. This method relies on an accurate prediction of local impact pressure. along the leading edge.1.1. 2. Numerically determine the local impact pressure.) fins (17) The body forces are well modeled with the current technique.1 Fin wave drag. fin forces are not well modeled. Pt2 . The accuracy of the current algorithm is shown in Appendix A.) body + (C. 3. Cd = (Cd. fin drag is not negligible [56]. but no correction was added to account for the wave drag created by the fins. by using the relation between fin sweep angle and impact pressure [64]. Therefore.4. 1. Assume that the pressure is constant across the flatfaced leading edge. + Cd. semiempirical corrections are proposed to obtain estimates of the fin components. Therefore.
.. C. The forces acting on a body by a fluid can be separated into pressure and viscous contributions. This is done with the Aeroballistic Research Facility Data Acquisition System (ARFDAS) [19]. They simplify considerably for a vehicle with fixed control surfaces. The analysis of multiple trajectories also gives better angle dependency. and mass properties are used to determine the aerodynamic coefficients and derivatives. position. The 6DOF routine incorporates the Maximum Likelihood Method (MLM) to match the theoretical trajectory to the experimentally measured trajectory.2.2.4. According to [35]. and still further for a bisymmetric one. are the surface pressures on the lower and upper surfaces of the fin just upstream of the fin's base. After missile models are fired through a test range.7 M )(.4. ) 2 where pL and p. The MLM is an iterative procedure that adjusts the aerodynamic coefficients to maximize a likelihood function. The theoretical equations of motion are provided by Etkin [18] for a general vehicle. it is assumed that pb! is constant across the base and (18) is used with with Pb! to determine drag. 24 . These data are passed to a routine that numerically integrates the six degreeoffreedom (6DOF) theoretical equations of motion. This program utilizes linear theory to get preliminary results. the use of this likelihood function eliminates the inherent assumption in least squares theory that the magnitude of the measurement noise must be consistent between dynamic parameters. ..4. the time history. An iteration process is used to match the theoretical trajectory to the measured trajectory.2 Pitchingmoment Coefficient. As above for the leading edge..twodimensional airfoils with flat bases [10].1 Experimental Methodology Using Trajectories and Six DegreeofFreedom Analysis. The following relation determines the nondimensional fin base pressure. pbf: Pb( Pb . (pi+.2 Computational Methodology Using Discrete Data. 2.G 2. 2. ARFDAS uses multiple flights in order to obtain a determination of Mach number dependence.
Solutions are calculated for a = 00 and a = 50.cosa+C. F...S. sina. = /s(rj~ifj . respectively. sin . and the resulting normalforce and moment coefficients are used to compute cno and Co. rTkifik'd. They are calculated here such that the axial direction is aligned with the xdirection. and the normal direction is aligned with the ydirection (see Figure 1).(9 F1= d§ (18) (19) • dS.: S=(a 5). (20) (21) The pitchingmoment coefficient is easily calculated from (18)(19) by including the appropriate local moment arm: M = js (Y)(Tjifij *d+ + (X.C(a = )(23) zAa 5")Cm(a ) Cm Cm ( (24) 25 . The axial and normal force coefficient are denoted by. Ci = CcosaC. derivative coefficients are derived from computation using steadystate discrete data and finitedifference expressions. and C.)hi'dS. C1.p.'dSD +r jT'dS(p .The decomposed forces acting in the idirection (xdirection in Figure 1) are S= /(pp. For finitevolume methodologies.) (. (22) Unlike the experiments. only two forces and one moment are nontrivial: the axial and normal forces and the pitching moment. For a bisymmetric computation. the projection of a surface area element into the three Cartesian directions. The aerodynamic forces can also be expressed in terms of lift and drag coefficients: Cd = C. d§ + Irkifik .) id).f(ppoo)iz. • dS relate directly to the surface grid. they are already known as the projections of grid cell interfaces (Section 3. and represent *j where fii • dS.. and fi.3)..
The properties of the boundary layer on the fin are deemed tertiary. a hierarchy of hii characteristics and phenomenon is used. 3.3. and fin crosssectional shape (thickness and leadingedge radius of curvature). The size.2 Description of the Thinfin Approximation The main reason for imposing an assumption about fin thickness is computational efficiency when studying complex flow structures. turbulent boundary layers and wakes. Secondary traits include: noslip fin surface. Previous computational efforts have been confined to the primary category described above. (11)(14). 1 Hierarchy of Effects To systematically analyze the stability of the HART missile. Highspeedmissile research requires the accurate simulation of shocks. fin proximity to base. and orientation of a fin will dominate its pitching capability. the philosophy of the thinfin analysis is discussed through a hierarchy of geometric and aerodynamic effects. with an infinitely thinfin assumption. 3. secondary. details regarding the approximation and the computational techniques are described. Following this general discussion. location. makes it possible in the present research to study the influence of many secondary fin traits and phenomenon. including turbulence modeling. Using the NavierStokes equations (Reynoldsaveraged for turbulent flow). First. Methodology for the Thinfin Investigation This chapter details the method used to investigate the HART missile with the thinfin approximation. These methods have used Euler equations. The fin traits are categorized here as primary. Only primary and secondary traits are modeled or simulated in this research. The primary characteristic is fin impermeability. and without a baseflow region. shape. amount of fin submerged in missile boundary layer. Finally. and shock/boundarylayer interactions. and tertiary. It typically takes many grid points to computationally analyze full configurations experiencing these phenomena. Turbulent transport of momentum influencing the flow in the finned region is also considered a secondary effect. the grid structure and boundary conditions are explained. The HART missile length makes it even more demanding to predict these structures accurately 26 .
The numerical implementation is discussed in Section 3. albeit without crosssectional thickness or curvature. this is expected to be a reasonable approximation [80]. Therefore. some special care must be taken when using them in the numerical algorithm. For fins which are very thin and highly swept. Since each point must represent at least two surfaces. any point coincident with the leading edge was handled differently than other points on the fin. and without extensive grid refinement.3 Computational Approach 3. A fin without thickness is quite straightforward to model.3.due to the disparity between length scales for a turbulent boundary layer and the length scale associated with the complete configuration. et al. One example is the Langley Aerothermodynamic Upwind Relaxation Algorithm (LAURA). Information at these points was extracted only from upstream coplanar data. each fin surface is a boundary embedded in the interior of the computational domain. The supposition is that dominant interactions in those directions can be calculated accurately without modeling the exact fin cross section.1 Overview of the Development of the Flowfield Solver.3. The first relates to the anticipated grid 27 . It simply is represented by a discrete set of grid points that are coplanar. but this was not the approach taken. for three reasons. In essence. This technique neglects effects due to the radius of curvature associated with the fin leading edge. at NASA Langley [25]. This is because the radius of curvature associated with the fin leadingedge is removed. Nonetheless. It has been used extensively on reentry problems. By assuming the fins to be infinitely thin. There are several solvers that were developed by other researchers to study full configurations in supersonic flow. the problem of extreme scale variation is encountered in all three dimensions. 3. including the Space Shuttle and the Aeroassisted Orbital Transfer Vehicle (AOTV) [23. Additionally. In general. and a limited number of calculations are done over a subdomain around the fin group. This is the subject of Part II where no approximation is made. Spatially coincident but distinct data are calculated on each fin surface. the leading edge does exist. These other techniques could have been modified for this work. 24]. the azimuthal grid requirements are dramatically reduced. This is a point implicit shockcapturing code constructed by Gnoffo. the flow solver is restrained from obtaining information through any fin.
requirements. The finest grids required over 1 million points. There is a very limited availability of resources capable of handling the memory requirements implicit solvers demand for this number of points. Therefore, an explicit solver was sought. Second, the validation computing requirement would have been much higher. As will be explained later, the tact taken was to develop an axisymmetric geometry solver first, then expand this solver for full threedimensional configurations. In this way, many numerical and grid related issues were settled at a reduced computational cost. Third, modification of an unfamiliar computer code without the benefit of inhouse corporate knowledge may not proceed smoothly and efficiently. This is especially true for codes that are many tens of thousands of lines long. Additionally, unanticipated anomalies would prove harder to diagnose and correct. Like other supersonic flowfield solvers for full configurations, the current technique had to be applied to a high Mach number and high Reynolds number combination, which is physically complicated and computationally demanding. The complexity arises from the development of strong curved shock waves, the associated vorticity which is generated, and interaction of this vorticity with the boundary layer. Turbulent transition and fin structures further complicate the interaction. Additional complexity is added by embedded subsonic flow that quickly expands to supersonic flow. Finally, the missile's flat base (see Figure 1) results in sudden flow separation, as well as a large mixing layer and recompression shock in the wake. Each of these phenomenon adds to the computational demands. Strong shocks and boundarylayer interactions make the correct amount of numerical dissipation uncertain. The subsonic flow involves a type change in the governing equations (from hyperbolic to elliptic), and turbulent boundary layers and mixing layers require very fine grid spacing to resolve steep gradients. Given this overall complexity, a computational methodology was developed that is sufficiently general to compute accurately a wide variety of flow structures. This methodology or procedure by which the governing equations, (11)(14), are solved is described in the following sections. The main aspects include threedimensional or axisymmetric equations, explicit time integration, timeaccurate calculations or steadystate calculations using local time stepping, and highaccuracy through the use of fluxdifference splitting and limiters. In addition, the algorithm chosen to meet the objective achieves secondorder spatial accuracy away from points of extrema (e.g., shocks). A practical bent was taken throughout the study; optimiz28
ing the computational speed was given a priority, and the memory requirement was reduced whenever possible. Some details are omitted here, and included in Appendix B. In particular, the eigenvectors obtained from a diagonalization of inviscid flux Jacobians are not included in this chapter, due to their length and complexity. The flux Jacobians come from the linearization of the nonlinear terms in (11). The construction of the grid and application of the boundary conditions are discussed after the algorithm is described. General Algorithm Description. The NavierStokes equations (Reynoldsaveraged for turbulent flows) are solved with an explicit, timeintegration technique, incorporating an upwind, Roetype, fluxdifferencesplitting (FDS) formulation based on the work of Harten [33] and Yee [104]. Because a finitevolume methodology is adopted, the integral form of (11) need not be recast into differential form. Equation (11) is repeated here for convenience. 3.3.2
"I
UdV+ fi.FdS=O.
(25)
This equation is a statement of the conservation laws for mass, momentum, and energy. In fact, use of a conservative form is necessary for shocks to represent physical waves when shockcapturing schemes are applied [1]. The fluid surrounding the body is partitioned into a structured, interconnecting set of cells, and the equations are applied to individual cells, incorporating the average flux over each face. For a fluid cell composed of 6 flat surfaces (Figure 10), this would take the following form using explicit time integration:
Un+1 =
A[
I(V, Fn)IV4 o)2
(g
)(
Fn)
The superscript n represents an arbitrary time level, and At is the time step from n to n + 1. The geometric terms §,(, §2c, S9, 97, 95, and g6 represent the projections of the six faces of the cell into the three Cartesian directions. The volume of the cell is V. The method used to calculate the geometric quantities from the grid points is 29
Figure 10. Typical fluid cell to which conservation laws are applied
adopted from Vinokur [91], and is outlined in Appendix B. For greater detail, the reader is referred to Vinokur's excellent treatment of the subject. Equation (26) yields information at the center of the cell, and the grid points are the vertices of the cell in Figure 10. The interconnecting cells are well ordered rather than random; this is usually referred to as a structured grid. The orderliness of the cell arrangement allows a mathematical transformation from the physical (z, y, z) to a generalized (f, 71, () coordinate system. Although a finitevolume method makes the mathematical manipulations unnecessary, the concept is important for the implementation of the FDS scheme. We can visualize this transformation with a box in computational space. In this box, the grid spacing is denoted by Af, Ai7 , and AC, such that ý = iAf, 71= jAq, and C = kA(. The node indices i, j, and k identify the cell center of interest. Equation (26) can be rewritten using these node indices as
30
Uijk= Uin u,+1
~j n, (S.tEn) , SE)+,•E)_, :E At,
k
(S

iLik
(
+ (svF),..k  (s(G),_, +
+ (S:E"),,+ 4k

)
(27)
(SI"),_Ik + (SF
n)i+,,, (s;F"),._k
+ (s~F),.÷+,  (s~F),,_k + (stG"),.k+  (SzG),.k_
First, consider the fluxes, E, F, and 0, in (27).
The explicit forms were
given by (12)(14). The convective terms are modified in an upwind manner, while the viscous terms remain unmodified. The modified fluxes are approximations to the average fluxes at each cell interface and, as such, are often called numerical fluzes. In addition, the modification is secondorder preserving for those interfaces located away from extrema, like shocks. These numerical fluxes can be determined at cell interfaces in two ways, using either MonotonicUpwindScalarConservationLaw (MUSCL) extrapolation or nonMUSCL extrapolation [104]. In the MUSCL approach, cellcentered values of primitive, conserved, or characteristic variables are extrapolated to the interfaces; numerical fluxes are then obtained from these quantities. The nonMUSCL approach, on the other hand, calculates the numerical fluxes at cell interfaces directly from cellcentered informatiL,n. The current scheme uses the nonMUSOL technique. NonMUSCL was chosen over MUSCL because it is more efficient a mre straightforward. The flux in the a coordinate direction at the (i +sc, k) interface, (reai "m )1 ' is therefore approximated by
tei avea=1
2 [ at
S'
(Ea+ljl
E,.,k) "t )+j (F +1,k +F+itrfcFand, 1
(28)
"+" (S•) i+½jk (G,+ljk +}G13 k) +FI (V1+ 11 k + Vjk ) ,li½,i½k]
The other interfaces have numerical fluxes with similar construction. Equation (27) is then written as
Ur'Uizi = U
1
t [E
+j 
. k + Fij+k
.. _½ +Gijk+½ . F+ Fi
1 • ½k (29)
31
These are calculated as follows: A 8E t s Sf . and the robust shockcapturing quality is preserved. FDS algorithm [71]. this nonCartesian method needs generalized inviscid flux Jacobians [104]. It is intended to characterize the scheme as a set of onedimensional operators. More detail can be found in [71] or [104]. shocks). The modification of the flux is responsible for the appearance of the final term in (28). which will be left unspecified momentarily. &V V V (30) 32 . Roe's idea is to perform a local decoupling of the nonlinear system with a linear wave decomposition.. which advances the solution of a nonlinear system of conservation laws to the next time level by solving a set of Riemann problems. and is the main mechanism used to prevent spurious oscillations near a shock. the global order of accuracy is not drastically degraded [104].g. The main feature of the method that makes it valuable for nonlinear systems is that it returns the exact solution whenever the left and right states lie on opposite sides of a shock wave or contact discontinuity [104]. global accuracy is increased from that for a firstorder upwind scheme. This behavior is common to all higherorder totalvariationdiminishing (TVD) schemes. and avoid the complexity of a truly multidimensional Riemann solver.TA + KB + LI. Roe's method was to approximate the nonlinear Riemann problem. the current scheme reduces to firstorder at points of extrema. To approximate the Riemann problem. This implies a localcharacteristic approach or generalization of Roe's approximate Riemann solver [104]. Thus. As stated above.The addition of the last term in (28) is a modification to Roe's firstorder. and solve the approximate problem exactly. Although the scheme switches to firstorder in regions of rapid change. The modified flux is cleverly chosen so that the scheme is secondorder accurate in regions of smoothness and firstorder at points of extrema (e. Secondorder spatial accuracy is achieved by applying Roe's firstorder algorithm to a modified inviscid flux [33]. He also requires that there exist an average state that is a nonlinear function of the states to the right and left [711. The directions over which the operators act are the generalized coordinate directions. This is a simpler approach than Godunov's method [261. Since Roe requires the diagonalization of the inviscid flux Jacobians.
evaluated at the (i + lj. proposed by Yee [104]. aG &= UV. A= B= . Since the purpose for this correction is to avoid the computation of nonphysical solutions. Due to the work of Harten and Yee. k) interface. andC=aG Returning to the final term in (28). $ also acts to limit the characteristic variables. 51. and simply acts as an upwinding term. as is well documented in the literature [32. $ also provides an entropy correction when the magnitudes of the eigenvalues become small. and a value proportional to the second derivative of the pressure. Entropy correction is not applied near the body surface to avoid adverse effects on the boundary layer. Three relations are tested in this study: a constant value. a value proportional to the contravariant velocities. In particular. Explicit forms for R and $ are also set forth in Appendix B. thereby providing higher accuracy. Other factors considered are impact on convergence to steadystate and the sideeffects of numerical dissipation. For Roe's scheme. The right and left eigenvectors appear through the diagonalization of A. put forth by Harten [34] is { I if JAI < f. supersonic. is denoted by R•+½jk. The form of the entropycorrected eigenvalues. $ is comprised of eigenvalues and characteristic variables. if (31) JAI. details regarding the development and implementation of the limiter are contained in Appendix B. for the current class of flows (external. 96]. the functions are evaluated on this basis. Lin [51] showed undesirable results when the function proposed by Yee was used in the boundary layer. The entropy correction prevents nonphysical solutions from developing. A C. A. Lin put forth a modification to 33 . suggested by Palmei and Ventkatapathy [63]. the matrix of right eigenvectors of A.where + E. proposed by Harten [34]. This is done in the current study with the minmod limiter. In this study.SE+ aE aF F+ SSG. where the proper value of c can be very sensitive to the geometry and the flowfield structure. axisymmetric nose) avoiding aberrations near the singular line is the primary concern.
is obtained when e is proportional to the contravariant velocities: + ICI). The calculations are performed at a Courant number of 0.9 and Re = 100. for the flows explored herein. results are judged best because nonphysical solutions were avoided.45. A value of 0. convergence is not impeded. Note that for both schemes the L1 norm matches 34 . For high Reynolds numbers.allow the function to be used globally. unsteady flows. without the addition of an entropy function. Driver [16] noted that secondorder temporal accuracy is not preserved for the NavierStekes equations with Strang time splitting. and shock smearing due to the added dissipation is minimized. the author feels that the natural viscosity of the fluid prevents nonphysical solutions near the body. For comparison. Figures 11a and 11b show the LInorm and Loonorm of computed errors for the current scheme and the secondorder scheme of Driver (ATNSC2) applied to the linear. Driver highlighted this fact for low Reynolds number. spatial and temporal for the Euler equations): . (IUol + IVcI + IWlI S= (32) where cl is varied from 0. the secondorder accurate line is plotted as the solid line and the firstorder accurate line is plotted as the dashed line.2 to 1.45 is used for all threedimensional calculations reported in this study. the most accurate and robust function. U +2 rh/2' h h hhhh/2 (jn (3 Lh J'Lh J( LhL~ :jk (33) where n L h/ 2 U tI 6ijk = =(ijk U =(ij ' i k * k ( j ~ (34) L hl h . By far. Utilizing the Strangtype fractionalstep [104] method ensures secondorder accuracy (spatial for NavierStokes equations.000. The best results are calculated with cl = 0. viscous. However. Burger's equation.0. This differs from the value used for axisymmetric calculations (Appendix A). * Application of the entire sequence of operators advances the solution two time levels. * U. the actual accuracy achieved is much closer to second order. c.
and Courant numbers close to 1._ik). when applied to separated flows.'.0.. Uj ( . As the grid spacing ig refined (Ax decreases). This is indicated by the Loonorm falling between the firstorder accurate line and secondorder accurate line in Figures 1l1a and 11b. indicating that the numerical solution is indeed globally secondorder accurate. the current scheme shows larger errors in both norms. note the firstorder (actually 1. (36) t =j (ij~k =(ijkAt (E'i  L:U. More details are included in Appendix A. because the inviscid and viscous fluxes within the numerical flux need not be recalculated for each operator.= .5) behavior in the Loonorm due to the TVD property. This limitation is overcome by evaluating the upwind terms for each operator with current information rather than timelevel n information. the norms for ATNSC2 decrease. made for Courant number equal to 1. therefore stability limitations of the current algorithm.+ 2 E k_ j  (37) = .ji).the secondorder curve extremely well. as required by (33) before. Ujk = Ujk At (pT+L P.•(•÷. the mild stability limitations of the current algorithm at high Reynolds numbers. this is much more efficient. secondorder algorithm developed by Driver is very expensive computationally. versus three and a half.0.. This fractionalstep method is (ijk+ I = rLL k (35) where again L.k  • • With nonMUSCL extrapolation. remain an open question.7J =i_. Additionally there are only three operators per iteration.. Since the explicit.. Also. The model problem does not have a region where viscous terms codominate. is an acceptable tradeoff. (Figures 12a and 12b) show that the stability behavior of the secondorder scheme is superior. In contrast. For steadystate calculations a much simpler approach is used. These same calculations. A stability limitation is encountered when using localtime stepping with this technique.. 35 .k but now LCT.
4. AX2] 36 . (b) (Re=100. .10 (a) 7 4D 10" O 0 0 0 10 o o 0 a .000.I Slope= 11 10. Accuracy of scheme when applied to linear viscous Burgers equation (a) current scheme. . AX2 ]. CFL = 0.9): ATNSC2 scheme. 0 [At 2. . 10' Ax Figure 11. 1" Ax o10" 10(b) AD 0 0 001 13 104 4 0 0 0li  o 2 SSlope SlopeI• 4 i0 . 0 [At. .
(b) ATNSC2 scheme. .0): (a) current scheme. . AX2 ] 37 .000. 0 [At 2 . 10"4 10• Ax (b) 104 . 0 0 0oo ¶0' 00 a SlopsI! Slope .2 io0. Accuracy of scheme when applied to linear viscous Burgers equation (Re=100.10 0 (a) z 0" 10 0 a 100 10 00 10' 00 0 ____  Slop I 2 S lop e = 10 '' 10' . Ax 2]. .. I 10"1 Ax Figure 12.. . 0 [A. 1 I. . CFL = 1. . 102 . 0  00 0 g z10. .
([13]. for the bluntnosetip model. (38) The nondimensional distance 26 is the incompressible value for a smooth. The effect of nosetip blunting to delay boundarylayer transition is documented in the literature. N is the pressuregradient contribution. Two modifications to the original BaldwinLomax model proved important. pressure gradients. 28. The original model is described in Appendix A.3. flat plate [101]. and widely used. and the wake. nonporous. In addition.1 Compressibility and Pressuregradient Modifications.25). zeroequation BaldwinLomax turbulence model [5] is incorporated for the forebody boundarylayer simulations. 101]. 3. this point is 6. for example) and agrees with the computations performed in Appendix A on a blunted ogive nose. No modifications are made to account for hypersonic flow. a twowall domain. save those for compressibility. All turbulent calculations have a point designated on the missile body to delineate the laminar and turbulent regions. Only important modifications are detailed in this section. This is also consistent with turbulent calculations made on other pointedogive noses in this study (Appendix A). Turbulence modeling is evaluated within the context of the assumptions made regarding secondary and tertiary effects (i. The near wall region.e. The limitations of this approach are not addressed as part of this research.4. For the sharpnosetip model. this point is selected on the ogive nose at (D = 1. is redefined as A+ = 26. denoted by A+.4 Turbulence Modeling The well known. and Z is the compressibility 38 . fin crosssectional shape is degenerate and boundarylayer separation is not anticipated. A point closer to the tip is selected for specified on the cylinder body at (= the sharp nosetip because early boundarylayer transition was seen in experimental shadowgraphs of the HART missile. for the thinfin investigation (Part I).. the boundary layer on the fin need not be accurately simulated).NZ.75). specifically those for compressibility. and therefore remain unresolved. The first changes the Van Driest damping factor to account for compressibility and pressure gradients [49.
1. The empirical technique was originally applied to subsonic flow by Magagnato [54]. The second alteration improves the two coefficients in the outer region.5. This modification has been implemented by other researchers [39. refers to the value of a variable at the wall. 73. Z .contribution: N [11. The grid limitations are discussed further in Section 3. The impact of this inadequacy is not clear. According to Granville.5).3 Empirical Model in Wake. CF and CkiLb. as reported by [55] for a dualwall region is di + d2 +A/ýf d d + 12 'd (41) where d. p+_ .4. The methodology of [55] is adopted without modification for this study. the following functional forms better match the law of the wake [27]: 2 3 0. 40] for wing/body junctions.2 Modification to the Length Scale in the Turbulence Model.: 3.1724 + p' 3  4Ckieb 2Ckieb (2 3Ckieb + Ckle)' dUe (40) (r*/p) dx y.(L P (dpe/dx) (39) [* ) ]W. The length scale for the turbulence model is locally modified to account for a twowall domain near the fins. 3. and d2 are the normal distances from the two walls. the boundary layers on the fins are not accurately simulated. The method is adapted and its validity tested for supersonic flow in 39 .? Z'. nearly stagnant region. The difficulty is that the appropriate length scale is hard to define and not readily computable. The length scale. e refers to the value of a variable at the edge of the boundary layer.4.n. and by McMaster and Shang [55] on a sweptback fin. k) IT *w) In (39).01312 0.8 ±e. A strictly empirical relation is used to compute turbulent viscosity in the wake because it is difficult to apply a conventional zeroequation turbulence model to a flat base with a large. P+_I i . Since the grid is relatively coarse in the azimuthal direction (see Section 3.1. and u.
The body shape (inner boundary) is generated analytically and input into GRIDGEN as a data file. are simple straight lines created within GRIDGEN. This facilitates boundarycondition application because the body and fin boundary conditions are most easily enforced for a grid constructed with gridlines normal to surfaces. Although highly empirical. For refined turbulent calculations it is reduced to 0. is not pursued in this study.this investigation. a single twodimensional grid is constructed using GRIDGEN. For this study. 3. The method has the advantage of speed and low memory requirement.5 Grid Structure All grids are obtained using GRIDGEN [79]. In this way. GRIDGEN has both twodimensional and threedimensional grid generating software. and & is the vorticity at the same point. The number of cells from the body to the outer domain varies from 81 for the baseline grid to 101 for refined calculations. Turbulent viscosity is determined from the local vorticity. The final two boundaries. and the fins coincide with distinct grid planes.2 (42) where At is the maximum viscosity in the boundary layer at the forebody/base junction computed from the BaldwinLomax model. and thus. and then rotated about the body centerline to produce a threedimensional grid. w. each plane is normal to the missile body. The number of grid cells and the appropriate axial and radial spacing were determined from the validation experiments (see Appendices A and B).000025D (y+ z 1). = 2 to fit within the domain at the exit plane. the validity of this approach is shown through numerical examples in Appendix A. As a result of these tests. The domain size (outer boundary) is chosen to allow a bow shock wave at M. an elliptical grid generator developed for Wright Laboratory. WrightPatterson AFB. A total of 91 cells are 40 . cells are clustered near the nosetip and fin/body junction. by 0.. Smoothing is not believed desirable nor physical by the author. singular line and outflow. Along the body. this spacing is 0. Ohio. the grids for investigating the HART missile are constructed with grid cells clustered near the body and with a fixed minimum cell height. Magagnato smoothed the eddyviscosity with an exponential damping factor.001D. For the baseline grid.
used in this direction. These are used to assess baseflow effects. and for an axisymmetric body at angle of attack. The value chosen above.05 (Figure 15). The minimum spacing near the missile body can be relaxed for the turbulent calculations with some accuracy degradation. and 31 in the wake. several other grids are used.053 and 518. For the sharpnosetip missile. and the latter a finitevolume (FV) grid. A nonrectangular domain composed of two regions results. These are summarized in Appendix A and 41 . In addition. The number of cells in the azimuthal direction and the angular spacing of these cells was determined from a refinement study discussed in Section 4. equal to 0. Cells can be constructed such that their centers fall on the singular line. A refined spacing for turbulent calculations is already mentioned. The modified nose is b = 0.2. and as such the prediction of skin friction. Two issues are important regarding the grid near the nosetip: tip geometry and singularline discretization. or such that the interfaces between cells lie on the singular line. A~min. these extremes represent 163. The baseline grid has cells clustered near fin surfaces with the angle between the fin and next plane of nodes.933 cells are used in the baseline grid (Figures 13 and 14). 0. 61 from the nosetip to the base. Calculations with each grid type are made for several axisymmetric flows. is required to obtain grid independence. Because the present research is concerned with corroborating the experimentally determined drag. one grid with 91 x 81 nodes and the other with 31 x 30 nodes. Larger spacing primarily affects only the velocity gradient very near the wall. the small spacing is selected. Both grid structures are shown in Figure 16. a refinement study for the azimuthal direction is made using equal angular spacing. In addition to the baseline grid. Dolling and Gray [131 found that spherically rounded with •/R 6 surface pressure (away from the tip) and boundarylayer development are virtually unaffected by this level of blunting. The number of azimuthal cells is varied from 33 to 105.000025D. This is more physical than an ideally sharp tip and enables the bow shock to be captured a small distance away from the surface. the nose is very slightly rounded. Without the base region. some grids are constructed to end at the missile body/base junction.805 cells. The former is called a finitedifference (FD) grid. respectively. A total of 273. Finally. The second aspect of the grid near the nosetip deals with the cell structure closest to the singular line.50 and 33 cells in the bisymmetric halfplane.
Baseline grid cutaway view : 60 x 81 x 33 around forebody and 31 x 111 x 33 in the wake 42 .Figure 13.
___ J43 .(b) 77.
.0 . I 0. a FV grid is used for all calculations in Part I. grid lines do not conform to the fin leadingedge shape. Modified nosetip showing very slight rounding of sharp nosetip Appendix B. . . The nonconformal nature of the grid represents an area of potential accuracy reduction because fluxes cannot be constructed such that they enforce the boundary conditions ezactly. the current a FV grid.3.0 1. the cells adjacent to the singular line degenerate into wedgeand prevent conserved quantities from fluxing into or out of portions of Although a FV method usually handle this without special treatment.0 0.0 0. . .5 x/D Figure 15.2) do not go to zero at the singular line.0 2. those cells.5 yID 0.5 1. .1. the fluxes are handled in a special manner near these points.5 2. For any interface that separates a cell upstream of a leading edge from one downstream 44 . Although grid cells are clustered at the fin/body junction. . However. To maximize the accuracy near boundaries. This is overcome by setting the entire numerical flux to zero. method requires some changes because the terms which are added to the physical fluxes to produce the numerical fluxes (Section 3.5 1. . . . 0.0 . the junction is a singularity that is not easily resolved with a single block as a structured grid. . For an infinitely thin fin. With like shapes.05 0.
.... (b) finitevolume 45 .(a) Singula Line (Axisynwmey Line for a = 00) (b) . Grid structure near singular line: (a) finitedifference.     $ingularlim (Axisymnwtry Line for a = 0°) Figure 16. L  .
6 Boundary Conditions The boundary conditions which were discussed briefly in Section 2. The grid is refined only normal to the missile body. The pressuregradient and adiabaticwall conditions are satisfied with secondorder discrete representations of: 9p 0. The author believes practical. the numerical flux is calculated from li =ti [ ()i+lj S)i+ijk G G(F1l1.of a leading edge. and . R. On 46 .5. v = 0.1 This is demonstrated for an Grid Limitations. the velocities are set to zero: u = 0. 3.&.Ik). 4 k(G' 1 Ij + Gi'+ljk) + Gtik) j (Vi+ljk + Vijk) Ri+j i+Ljk] . and ( + instead of Uijk. 3. Since the pressure is dominated by the shock structure. and w = 0. Again. albiet not formal. it is emphasized that the azimuthal refine ment is insufficient to predict accurately the boundarylayer properties on the fin surface. and the boundary layer of interest is the one on the body. For the noslip condition used with viscous calculations. 1 where u and 1 represent fluxes calculated on the upper side and lower side of the fin respectively. and Ui+2 .3 are now described in detail.k + E~j + Ei+ljk) + Ei 3k) + Fi+ljk) + Fijk) (43) + (~). this limitation is not deemed overly constraining by the author.n =p O.+½jk and 'i+½lkare now functions of Uijk. This will be addressed in detail in the thickfin analysis (Part II). accuracy is preserved. Also. 2 (Ui+ljk + U•+.•k. axisymmetric compression ramp in Appendix B. Ui+.
the condition is not so simple. First. (46) where cl and c2 are constants that depend on the flowfield.e. Finally. 3.0 can be determined. the impermeability condition for a threedimensional surface is discussed. = 2 qr• (('r) 8•2qI (r+ (Ar)O+ ý2ear 2r 2 q2 + + +0[(Ar)3] . except v and w.0). the condition at the singular line is a symmetry condition. and q4 . The variable q represents any primitive variable. When a body is not pitched (a = 00). The remaining velocity components. For the current FV grid.1 Singularline Conditions. v and w.Three of the boundary conditions require further explanation. By using (46)(47) to determine q2. behave as the radial velocity (except for a sign change) [75]: q • car (r . When a body is pitched. q3. the ext rapolation at the singular line is described. Assuming that the grid is constructed with a nearly constant node spacing normal to the 47 . a cell center slightly displaced from the symmetry line is mirrored by one displaced exactly the same amount. (44) The subscript 2 denotes the first cell on the domain interior. For axisymmetric flow.6. It is a simple matter to reflect v and w to achieve axisymmetry. the behavior of q.. (47) where c3 is also a constant that depends on the flowfield. but azimuthally located 1800 apart (Figure 16). Next. behave like [75] q . (45) It is important to check the consistency of this extrapolation as a + 0. Therefore. the finsurface conditions are shown. The subscript 1 denotes its mirror (i. the boundary cell). = 3q 2 . all primitive variables.0). Using finitedifference expressions for the derivative terms in (44) yields q. as r . the following extrapolation is used whenever a #' 00: q.3q3 + q4 + 0 [(Ar)3] .C1 + c 2r 2 (r .
and swirl velocities (un. 2 (50) and again from (45) q 1 =cI + C2( 2 fAr\) = q2.cos0 0 sin4. ut. 3 q4 : 5Ar c3  . C 3 . u. an impermeability condition is required. but for inviscid calculations. [u] =[L] v w cos 0 cos4' sin 0 = sin0 cos4. In a similar manner. the local body inclination.6. as r . sin 0sin 0 sin4'cos0 cos 1u] v j (52) w V =[LF W 48 U4 ]t (53) . no slip is enforced on the missile surface. 3. (51) For a grid with constant spacing normal to the singular line.symmetry line. tangential. = c q2 CI + C2 (). the proposed extrapolation is consistent with axisymmetric flow. 0.2 Impermeability Condition for a Threedimensional Surface. q3 CI + C2(CI2q 4 +C().= q2. and u. and the azimuth. q3 :  0. then for v and w. (Figure 17). Therefore. For viscous calculations. 0. (48) and from (45) Ar (49) qI 3 . and p. 4. for p. 3Ar C 2. as r Ar q2 . 2 The distance between the cell center at 2 and the symmetry line is given by Ar/2. The impermeability of a threedimensional surface depends on two angles. respectively): [n ] Uj u. A simple coordinate transformation using these two angles produces a relation between the Cartesian velocities and the normal.
ffi 2 . 0. v w Finally. and 03 = cos 0 sin 0.. . and (u.=f2 . the tangential and swirl velocities are specified with the following: (Ut). 0.(u. very little memory is required to keep track of 4) and 0.)j=l = (un).= 2 .sin2 0. Body orientation showing references for body inclination angle. 49 . and azimuthal angle. Thus.r . #2 = cos 0 sin4.sin 2 4++cos2 S2 J w Jj= 2 01= cos2 0 . the angles are calculated as functions dependent on only one grid index: 0 = 4(k) and 0 = O(i). where y = r cos 0 and z = r sin 0 For impermeability. In addition.). In fact. and the grid is rotationally generated.= 1 = (Ut).). the normal component of velocity is specified to vanish at the missile surface: (u.X Figure 17.=. Because the HART missile body is rotationally symmetric. u = jfilj2 u us 0 0 1 0 0 1 [L] v w j2 (54) ][ 1 = 2#3 cos4) 2 /3cos 4)+sin 24) 2#3 sin 32sin4)cos 0 /31/ 1u] vI (55) v w j= where 2#cos4)' 2#3isinO 61 #i2 sinocosO #.
depending on which side of the fin u. Using dual variables.6. The Cartesian velocities are transformed into cylindrical velocities to facilitate the process: ["u 1 =[L] u. (UG)k=kfin = (Ua)kfkfin. and 02 is the azimuthal angle of a point above or below the fin. The mechanics of this condition are different from Section 3. (uo)k=kfi= 0.sin• 0cosb For impermeability. For the body surface. the vanishing pressuregradient and the adiabaticwall conditions are enforced from above and below. Since the fins have no thickness. but for the fin surfaces._1. jw v 0 0 V w (56) coS4€ sin 0 . Thus. The main difference is the location where impermeability is enforced.3. Similar expressions are Also. and w=0. u. = and (UT)k==kin lU of each fin. and w are desired. it is at cell centers. v=0._1. This means that the swirl velocity is set to zero at a cell center.6. (uT)kknl'. For viscous calculations. Additional variables are introduced to accomplish this. on the top used for the bottom of each fin. v. u = v wJ k=kfin 1 0 0 u1 v W cos 01 sin 0 2 0 cos •b 1 cosS 2 0 sin 01 COS 02 sin 01l sin 0b2 1 (57) =kfin+l where 01 is the azimuthal angle of a point on the fin. and Surface Conditions.2.3 Fin Leadingedge. Trailingedge. it is between cells. As with the missile surface. no slip is also enforced on both sides of a fin. 50 . the same set of points in the grid that represent the upper surface of a fin must represent the lower surface. Again this is given by u=0. fin surfaces in inviscid flow require an impermeability condition.
Table 1 lists the computations made to assess two sensitivity issues.5000 15.5 3.8125 1.5 3.5000 . All sensitivity calculations are conducted at M. The computations are summarized in the following paragraphs.001 0.0 5.5000 0.6250 2. Results and Analysis from Thinfin Tests The results from this study are presented in terms of a gridsensitivity analysis.5000 15. (Background for these results can be found in Section 1.5 0.0 5. The spacing is varied from approximately 6" to 0.) 4. 93] serve as the primary basis for comparison. These include calculations to determine grid sensitivity. Table 1.c R1 R2 R3 R4 R5 R6 3.7308 15.50 (Cases R4R6) to determine the variation of Cm with a. The experimental results and the EAGLE computational results [22. Summary of numerical experiments on the bluntnosetip model used for grid sensitivity analysis I xJ x K 5.5 3. and baseflow effects.001 0. In addition. a flowstructure analysis. the angle of attack is varied from 50 to 0.5*.001 laminar laminar laminar laminar laminar laminar The case history of calculations made for the sharpnosetip model.001 0.5 &OAMax 5. a drag analysis.5000 [Case] M.1 Summary of Numerical Ezperiments Over 80 numerical computations are performed using the thinfin assumption.5 3. and a staticstability analysis.1. aerodynamic characteristics for the sharpnosetip and bluntnosetip models.4.8125 1.0 5.0 2. = 3. see Section 3.001 0.7308 0. is tabulated in Table 2.5. without the base region.5000 0.6). The Mach number is varied from 2 to 51 .6250 2.5 3.5 61 x 81 x 33 61 x 81 x 65 61 x 81 x 105 61 x 81 x 33 61 x 81 x 33 61 x 81 x 33 Abmin 5.001 0.bodyJComments 0. Cases R1R4 vary the angular spacing (AO) near the fin to determine the sensitivity of the axialforce coefficient and the pitchingmoment coefficient to the boundary condition that is used to extract pressure on the fin surfaces (vanishing pressure gradient.
the axial force does not depend on node clustering in the azimuthal direction... Ca varies only 1%.00005D.. only two laminar and two turbulent tests are conductedwith refined spacing near the missile body. S12. laminar. Cases S11. Due to limited resources. computations with the refined grid are done for only two Mach numbers. Mach number is varied from 2 to 6 and calculations are made for a = 00 and a = 5°. and the grid spacing near the missile body is held fixed at 0.. two cases involve a larger fin. shows a much greater sensitivity to anb. Figure 19 demonstrates that C. and S17 were computed by Blake [8] as part of a Master's thesis involving parallel computing techniques. S33. Also. Like the sharpnosetip tests. In addition. The pitchingmoment coefficient. C. calculations that include the base region are also made. Only turbulent calculations are performed. Computations which include base effects are designated with an "SB'W (Table 3). this spacing is 0. Also.5. Mach number is varied from 2 to 6. Also. and S34). Over the range of spacing tested (6* to 0. S16. and turbulent flowfield conditions.00005D. Ca. Again. One case involves a larger fin. by the gular spacing. For the sharpnosetip model. The aerodynamics of the bluntnosetip model are also determined with a parametric Mach number study. The summary of cases for the bluntnosetip configuration is provided in Table 4.• 1. and the effect of angle of attack on base drag. The spacing near the missile body for these cases is equivalent to y+ . 4.1. A limited number of computations are performed with refined spacing near the missile body (Cases S21. and calculations are made at a = 00 and a = 50 for inviscid. S22. For the bluntnosetip. they are used to study the impact of the base region on pitching moment. is influenced signific 52 .5.6. and are explained in Section 4.2 Grid Refinement and Sensitivity Figure 18 shows the axialforce coefficient plotted against the angular spacing near the fins. is found to be very insensitive to the spacing near the fin. respectively).. Blake used the present algorithm to make those computations.5*).1. The axialforce coefficient. and is explained in Section 4. as with the sharpnosetip study. This is evidenced by the nearly identical results for unclustered and clustered spacing applied to the same number of azimuthal nodes (Cases R1 and R4. These are used to assess the influence of the fins on base drag.
Table 2.001 0.0 3.001 0.001 0.001 0.001 0.001 Inviscid (coarse grid) Laminar (coarse grid) Laminar (fine grid) Turbulent (coarse grid) Turbulent (fine grid) Large fin 53 . Summary of numerical experiments for sharpnosetip model.001 0.0 3.0 2.0 2.5 4.001 0.001 0.5 3.0 3.0 3.5 4.0 3.001 0.5 4.5 6.001 0.001 0.001 0.0 2.000025 0.001 0.000025 0.001 0.0 2.001 0.001 0.001 0.5 6.5 4.5 4.001 0.001 0.0 3.001 0.0 3.0 2.0 3.5 2.5 4. without base region Case Si S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S20 S21 S22 S23 S24 S25 S26 S27 S28 S29 S30 S31 S32 533 S34 S35 S36 M 2.001 0.000025 0.001 0.001 0.0 2.5 6.0 3.001 0.5 6.001 0.001 0.0 3.0 3.5 a 0 0 0 0 0 5 5 5 5 5 0 0 0 0 0 5 5 5 5 5 0 0 0 0 0 0 0 5 5 5 5 5 0 0 5 5 IxJxK 61 x81 x33 61 x81 x33 61 x81 x33 61 x81 x33 61 x 81 x 33 61 x 81 x 33 61x81x33 61 x81 x33 61x81x33 61x81 x33 61x81 x33 61x81 x33 61x81 x33 61x81x33 61 x 81 x 33 61 x 81 x 33 61 x81 x33 61x81x33 61x81x33 61 x81 x33 61 x 101 x 33 61 x 101 x 33 61 x81x33 81 x 33 81 x 33 u x81 x33 61 x 81 x 33 61 x 81 x 33 61x81x33 61x81x33 61x81x33 61x81x33 61 x 101 x 33 61 x 101 x 33 61 x 81 x 33 61x81x33 0.001 0.001 0.0 3.001 0.001 0.5 4.0 3.0 3.5 6.5 6.000025 0.001 0.
near the fin 50 40 C_ 30 lauem . . I 0 1 2 3 4 5 6 Figure 19. to angular spacing. 0. A#at fin(degrme) Sensitivity of pitchingmoment coefficient.30 . ..u.90 0. I .50 0.I .I 0 1 2 3 4 5 6 Figure 18.a.40  I aum Zimuthal I 0.80 0.. AO. .im. C.0. AO. near the fin 54 .spcinl 20 . Cm. A* at fin (degree) Sensitivity of axialforce coefficient.70 C. .. to angular spacing.60 0.
Thus..00005 0.. and only over a portion of some trajectories.. with base region Case SBI SB2 SB3 Turbulent SB4 SB5 SB6 SB7 1 SB8 Large fin SB9 M_____ 2.5*). As the angle of attack increases.07 0.00005 0.0 5 I i(Forebody) 60x81x33 60x81 x33 60x81x33 60 x 81 x 33 60x81 x33 60x81 x33 60x81x33 60x81x33 60 x 81 x 33 Io I x J xK I XJ XK (Wake) J "/5 b d b Jyi 'AX e J 31x111x33 31x 111 x33 31x111x33 31 x 111 x 33 31 xiii x33 31 x111 x33 31x111x33 31 x111 x33 31 x111 x 33 0. grids with uniform angular spacing less than 1. In addition. The rate at which the pitching moment vanishes as a + 0 is shown in Figure 20.07 0. Unfortunately. As AOb decreases from (6* to 0.. other boundary conditions to obtain pressure were not attempted.. decreases (more negative) about 30%.00005 0.0 5 2.2 to 8. the HART missile occasionally achieved angles of attack in excess of 5 degrees.5 0 6.0.. the use of node clustering near fins overcomes the deficiency of the pressuregradient boundary condition. 55 .0 5 3. Therefore. resource limitations prevent this analysis from being repeated at other Mach numbers.00005 0. found by Riner is nearly equivalent in magnitude to the difference in C. from the present computations between the coarsest and finest spacing used. Due to resource limitations... Summary of numerical experiments for sharpnosetip model..0 0 3.00005 0.07 0. appears to be approaching an asymptotic value near 0. he found that the pressuregradient boundary condition underpredicted C.70 (Case R3) were not attempted.07 0.Table 3..00005 0. Also.07 0.. Riner [69] found that EAGLE computations made with the pressuregradient boundary condition and characteristicvariable boundary conditions differed dramatically.5 5 6.. and nodes are clustered near the fins throughout the rest of this study. plotted in Figure 20 differ by about 1%. the pitchingmoment coefficient changes only slightly. The three values of C.. is independent of a. C. the results at Mo. However. C.07 0... = 3. In freeflight testing. For a grid with constant angular spacing and AO = 10".0 0 2.07 angular spacing near the fins.07 0.5 5 4.00005 0.00005 0. by a nearly constant amount over a Mach number range from 1. which indicates that C.07 0.5 agree with inviscid linear theory given by (9).00005 0.5*.5 0 4. However. The difference in Cm.
001 0.001 0.001 0.5 5 B20 5.001 0. Summary of numerical experiments for bluntnosetip model.Table 4.001 0.0 5 B22 2.001 0.0 0 B7 6.0 0 B8 2.95 0 B31 3.0 5 B29 6.001 0.5 0 B24 2.001 0.001 0.001 0.001 0.001 0.95 5 B10 3.5 5 B27 4.0 5 B25 2.0 5 B9 2.001 0.95 0 B3 3.001 0.001 0.001 0.5 0 B6 5.5 5 B13 5.95 0 B23 3.0 5 B21 6.95 5 B17 3.0 5 B19 4.0 0 B2 2.001 0.00005 0.001 0.001 0.00005 0.001 0.0 0 B5 4. I a ~ L II B1 2.0 5 BI5 2.001 0.001 0.95 5 B26 3.0 5 B16 2.0 5 B28 5.0 5 B12 4.001 0.001 0.0 5 B30 2.5 5 B11 4.001 0.001 0.00005 0. without base region Case I M.5 0 B4 4.00005 56 .0 5 B14 6.5 5 B18 4.5 0 I x JxK 61 x 81 X'33 61 x81 x33 61x81 x33 61x81x33 61 x81 x33 61 x81 x33 61 x 81 x 33 61 x 81 x 33 61x81x33 61 x81x33 61x81x33 61x81 x33 61x81x33 61 x81 x33 61 x81 x33 61x81x33 61 x 81 x 33 61 x 81 x 33 61x81x33 61 x81 x33 61 x81 x33 61 x 81 x 33 61 x 81 x 33 61x81x33 61x81 x33 61 x 81 x 33 61 x 81 x 33 61 x81 x33 61 x81 x33 61 x 81 x 33 61 x 81 x 33 _______ I bodv Inviscid (coarse grid) Laminar (coarse grid) Laminar (fine grid) Turbulent (coarse grid) Turbulent (fine grid) 0.001 0.
C.50 40 C 30 20 I . The wedge angle is equivalent to the angle of attack of the thin fin. In Figure 21.3. a = 50 is representative of the experiments and is selected in the current study for all remaining computations performed with a nonzero angle of attack. to angle of attack.3 Flowfield Near Fin The flowfield near the fins is examined through analysis of the shockwave shape and strength of the shock/boundarylayer interaction.. . 4. For the delta wings. 4. The interaction of the shocks with the boundary layer growing on the missile body is assessed with surface pressure line plots and contour plots. 0 1 2 3 4 5 6 a (degrms) Figure 20.1 Shockwave Shape. The angles of the shock waves created by the fins are compared to the theoretical and experimental angles discussed in Section 2. Sensitivity of pitchingmoment coefficient. small wedge angles and highly swept leading edges produce shocks that emanate from the leading edge at angles very close to the Mach angle. the shock wave created by a thinswept fin at angle of attack is compared directly to the shock wave from a wedgeshaped delta wing.. The Mach angle based on the local Mach number is indicated with a dashed line in 57 .1.1.
2 Shock/Boundarylayer Interaction. The pressure ahead of each fin 58 . Figure 21. the compression on the lower surface and the expansion on the upper surface are distinctly separated in the inviscid flow. laminar flow (Case S16). Note the clustering of contour lines near the leading edges in Figures 22a and 23a. attached shocks are seen on the lower surface leading edges. pressure is plotted along the missile body upstream of the fin/body junction for Mo = 2. and indicates that the shock/boundarylayer interaction is weak. Figure 21 shows pressure contours on the missile body for Moo = 2 and a = 5* (Case B8). The pressure contours show a shock below each fin and an expansion above each fin. as evidenced by the clustered contour lines in Figures 22b and 23b. = 2 and a = 5°. Likewise. The boundarylayer does not separate in any of the laminar or turbulent computations. In Figure 24. a = 5°. To further assess the weakness of the interaction. Although the pressure contours show weak compressions on tbe upper fin surfaces and weak expansions on the lower fin surfaces. Rapid expansions to the upper surfaces are visible at the leading edges of both fins.Figure 21 for reference.3. and turbulent flow (Case S28). Pressure contours on the missile near the fin/body junction (Case 138) 4. In addition. the pressure is relatively constant on each surface. Figures 22 and 23 show the pressure on the upper and lower surfaces of the two fins (one on the expansion side of the missile and one on the compression side of the missile) for M. the surface pressure is examined. The shockwave angle is well predicted using the thinfin assumption.
5 22.50 A 9 8 7 6 5 3 0.1571 0.2265 02149 92 r 125 1.1339 23.1455 0.0 .0 21.75 0. 21.0 X 22.00 0.a 21.25 Figure 22.75 0..1339 r 1.1802 .5 23.5 2 1 (b) 0.0 S5 0.0 2.1802 0.5 23.5 22. Pressure contours for the fin on expansion side of the missile (Case B8): (a) upper surface.00 (5) A 1.1918 0.0 23. (b) lower surface 59 .2361 0.75 1.2.50 9 8 7 02381 0.75 1.2034 0. ••__• •l•3 I1 4 2 018 015 0.2265 02149 0. 21.00 1.1571 0.50.5 x 22.50.
0 2.75 1.00 1.5 22.1571 . Pressure contours for the fin on the compression side of the missile (Case B8): (a) upper surface.1916 0. 0.0 Figure 23.50 A 9 8 7 (b) 02361 02265 02149 02034 r1.0 22.25 r•• 1.5 22.0 A 9 8 7 (A) 02361 02265 02149 02034 r 1.1.00 1.0 21.5 24.75 0 0.1802 0.5 0.1455 0.0 2 1 23.5 x 23.2.5 x 23. (b) lower surface 60 .0 23.005 0.0 21.0 22.00 Ole_ ?m m • 7 5 6 5 4 3 olo 0.50 21.1571 0.75 1.501 21.1339 24.25 1.1687 0.75 4 47 3• 2 0.
the influence of the fins on the base drag and the effect of angle of attack on the base drag are discussed with plots showing the variation of base drag with Mach number. the drag obtained from freeflight testing is compared to the current computation of total drag. First. All drag values computed in this study at a = 0* are summarized in Table 5 for the sharpnosetip model and Table 6 for the bluntnosetip model. but other researchers have shown that highlyswept fins exhibit no boundarylayer separation [64. Moreover. and is nearly the saine for laminar and turbulent conditions. Finally. it is directly obtained from calculations on the complete HART missile configuration. The pressure is only slightly disturbed ahead of the leading edges. Comparisons with freeflight data are also made with Cd versus Moo plots. the drag coefficient is broken down into components. The wavedrag coefficient and pitchingmoment coefficient from EAGLE computations on the sharpnosetip model with clippeddelta fins are summarized in Table 7.4 Drag Analysis To facilitate a detailed analysis of missile drag.is plotted separately. this is due to the weakness of the shock/boundarylayer interaction. The drag breakdown was discussed in Section 2. The following sections discuss and analyze each component more fully. Eliminating the fin's cross section may have prevented separation. Additionally. 99. The base drag is also computed for axisymmetric bases. In addition. These simpler computations allow much finer grid resolution. The coefficients are tabulated for Mach numbers ranging from 2 to 8. basedrag. 55]. 4. Total drag is computed using the base drag from Appendix A because of the greater level of validation. the plateau associated with separation [2] is not visible. laminar. These tables contain the wavedrag. and frictiondrag c(. Again. Included 61 . The base drag in the current study is computed in two ways.1. using axisymmetric equations. The empirical components used with previous EAGLE computations are graphically compared to quantities directly calculated in this study.4. and are described in Appendix A. and the components are analyzed individually and then collectively. and the location of the fin leading edge is indicated for reference.3onents for the missile body and the fins tabulated separately. or turbulent) and grid coarseness. the data are delineated by the flowfield condition (inviscid. in axisymmetric flow (a = 00).
.. 0.80 I 20..25 .25 20....00 1.75 I 21....000 p .050 ...75 21.80 I 20.. Pressure along the missile body upstream of the fin/body junction: (a) laminar flow (Case S16)..I .025 P 1.025 P1.1.   FI IFm Le.00 Figure 24.. x5 x 20. . 7 .  0.060 .000 p.975 0.. (b)turbulent flow (Case S28) 62 .00 20.  =45' *a135" 1.. koradio 0.950 20.00 20. # 45" 4135" 1..975 .950 I 20.
357 0.049 0.Table 5.137 0.079' 0.204 0.334 0.008 0.162 0.160 0.045" 0.047 0.257 Turbulent 0.145" 0.137 0..468 0.281 0.00) ICase K1 (coarse grid) Cd I(bo 0.2.042 0.010 0.145" 0. and shows very little sensitivity to viscous effects.145* 0.043 0.079" 0.165 (fine grid) S34 SB1 0.244 0.047 0.129 0.043 0.1.007 0.131 0.044 0.044 0.265 0.145 Laminar 0.013 0.042 0.268 0. and turbulent calculations.093 0.053 (fine grid) 0.140 S3 S4 0.Base drag taken from Appendix A.007 0. laminar.012 0.140 0. or wave drag.416 0.034 0. fixed.611 0. where the wave drag is nearly identical for inviscid.147 Turbulent SB2 0.142 0.048 0.059 S27 S33 0.162 0.007 0.024' 0.049 0.010 0.024' 0.147 0.009 0.129 S11 S12 S13 S14 S15 S21 S22 S23 S24 S25 S26 J(bocy) (bo ly)  Cd Cd Cd + fin (fins)CdblinC  I tota) C 0.045 0.043 0.142 0.0450 0.162 0.160.021 0.103 (with base) SB3 0.046 0.006 * 0.043 0.079' 0.046 0. The following sections compare this data to the present results.079* 0.438 0.1 The wave drag from the turbulent calculations on the sharpnosetip model (Cases S23S27 and S33S34) is plotted versus Mach number with the EAGLE results in Figure 25.599 0. in Table 7 are the empirical values for viscous drag (turbulent) and base drag used to correct the inviscid EAGLE results.262   0.042 0.024" 0. 0. Also included is the value predicted by slenderbody theory [4] 63 .051 0.047 0.042 0.078 0. 4.145" 0.054 0:140 (coarse grid) 0.048 0.060 1_SB4 .160 Laminar 0. and are given by (1) and (2).352 0. is dominated by the nosetip geometry.619 0.079' 0.008 0.433 0.010 0.006 0.145' 0.012 0.242 0.044 0.098* 0.006 0.008 0.407 0.291 0. The forebody pressure.074 0.098" 0.009 0.135 (coarse grid) 0.131 0.327 0..045' 0.057 0.151 0.045 0.050 0. Computed drag components for sharpnosetip model (a .135 S5 0.4. Forebody Wave Drag. This is seen in Table 5 or Table 6 with M.007 0.434 0.145 S2 Snviscid 0.045 0.098" 0.326 0.073 0.168 Turbulent 0.135 0.328 0.238 0. These corrections were detailed in Section 1.142 0.230 0.
016 0.574 0.619 0.086 0.492 0.286 0.457 0.130 0.019 0.Table 6.145 0.. = 4 Laminar M.484 15.529 0.100* 0.558 0.457 0.475 0.043 0.99 0.073 0.108 0. = 3 M.92 0.568 0.720 0.063 0.661 62.022 10.0  Q I Cd [Cm Q.475 0. Computed drag components for bluntnosetip model (a = 00) Case BI B2 B3 Inviscid B4 (coarse grid) B5 B6 B7 B22 Laminar B23 (fine grid) B30 Turbulent B31 (fine grid) Mo.035 0.025 0.584 0.465 0.024 0.110 0.078 0.41 0.121 0.145* 0.0 3.126 0.060 0.423 2.467 11.028 0. = 6 Moo = 7 * .036 0. Cd (bocy) 0.023 0.079* 0.21 0.023 0.580 0.Base drag taken from **  Cd [C (boy) (body) 0.. = 2 M.457 0.490 0.272 0.415 0.563 39. 0 ] 0.100* 0. Table 7.079* 0.0790.100* 0.552 0.310 0.539 0. 2.129 0.019   0.441 5.671 0. 64 .018 0.30 0.465 0.330 0.024 0.159 0.025 0.136 0.409 **  Empirical.063* 0.015 0.039* 0.011 I Cdf +SCdb I Cd 4 I Cd (fins) (fins) (total) 0.465 0.0 8.296 0.024 0. = 5 (with base) (AxiEqns)** M.83 0.010  0.0 4.541 0.659 0.477 0.416 0.015 0.022 0.528 0.278 0.045* 0.88 0.266 0.037 0.024* 0. Empirical.010 0.479 0.484 0.0 4. Drag and pitchingmoment coefficients from EAGLE results for sharpnosetip model with clippeddelta fins [Moo I Cd. also corrected to account for Reynolds number error..051 0.024 0.028 0.471 0.172 0.024 0.453 0.09 0.0 6.291 0.133 0.021 0.544 0.523 0.505 22.049 0.5 5.015 Appendix A.123 0.519 Calculated from axisymmetric equations for bluntnosetip model without fins.0 7.
This is also seen seen in Figure 3 for the baseline HART model. The variation with Mach number is again small. This was done for previous EAGLE results (Table 7). (2). however. The trend in Cd.and the arearule theory [15] for Moo = 2 (see Section 2. the variation with Mach number is modest. = 2 to 0. the turbulent friction drag from the coarsegrid test (Case S23) is 0. Cd.135 to 0. The wave drag from the bluntnosetip model is plotted versus Mach number in Figure 26. decreases from 0. At M. = 8 compressib~lity results in a very large decrease in Cd.5 to Mo..238. increases. By this method. the wetted area of the HART missile is calculated and used with an empirical relation for friction drag on a flat plate to compute friction drag for the missile. as Mo. except slenderbody theory. using an incompressible formula.131 at Moo = 6. the laminar friction drag is calculated equally well with the coarse grid (Cases $11$15) and the fine grid (Cases S21S22). The effects of compressibility on friction drag cannot be understated. One method to compute friction drag is to utilize the concept of a flat plate of equivalent size.126 at Mo.. For Mo.257. The relatively flat profile of the relation used to correct the inviscid EAGLE results is a direct consequence of neglecting compressibility effects. The agreement between all the results. Cd. 65 . The wave drag from the current coarse grid and fine grid computations is identical. with increasing Mach number from the compressible flatplate formula agrees well with the current computations. A similar procedure is employed here with a compressibility correction to (2). Because the velocity gradients near the wall are less severe for laminar flow than for turbulent flow. and both equivalent flatplate formulas. Likewise. = 3. Accurate prediction of friction drag depends primarily on adequate grid resolution normal to the missile body surface. decreases from 0. 4. is considered excellent. An accurate prediction of the turbulent friction drag required more refined grid spacing near the missile body. As expected from theory. and turbulent friction drag from the finegrid test (Case S33) is 0. at M. For the present computations.165. = 2. Turbulent skin friction on a flat plate varies considerably with supersonic Mach numbers [101].2 Forebody Friction Drag.162 at Moo = 2 to 0. Here the wave drag increases with Mach number. Figure 27 compares the turbulent viscous drag (Cases S23S27). = 6.159 at Mo. = 1. the turbulent friction drag increased with grid refinement from 0.1). For the EAGLE computations.4. the EAGLE results.2.5.
. . Cal. 0. for bluntnosetip model (a = 00) 66 . Wavedrag coefficient.0. Cd.55 C. 1 & 0..30 o. 7 8 Figure 26..10 .65 O.. for sharpnosetip model (a = 00) 0.45 0. Wavedrag coefficient. 1 2 3 4 5 6 7 8 Figure 25.40 0..lO 0. I. 6 ..70 0. 5 .15 .50 0..35 0.25  EAGLE Curren (warse gri) slanderSid ftc~v Eauatic (9) V ArearulemUay: Rdarive [15 C 0.30 1 2 3 4 M.
The base drag for a = 50 increases slightly. Love showed that both the proximity of the fins to the base and the thickness of the fins can alter base pressure. this demonstrates the accuracy that can be achieved with the empirical formula. The reevaluation was done through an existing contract between Wright Laboratory and Arrow Tech Associates. The curves cross around M. this author initiated a reevaluation of the freeflight data in order to quantify the error in the trajectory match for each test and obtain more reliable data.. 4. = 3 due to the effects of the fins. Base drag should increase with increasing angle of attack [76]. He also showed that the interference effect decreases significantly with increasing Mach number. the drag from the complete configuration with fins is greater. The empirical correction used with the EAGLE results. At the lcwest Mach number tested. The base drag from the complete configuration calculations is generally lower than the axisymmetric calculations. Very similar results are obtained for the two angles of attack. and is documented in a contractor test report [37]. The correlation is good for Mach number from 1. The empirical formula consistently predicts a higher value of base drag for all Mach numbers. The reason why it was more difficult to correlate trajectories for the clippeddeltafin was not ascertained. The original and reevaluated drag data are plotted in Figure 31.. (1). is compared to the values computed with the current axisymmetric equations in Figure 28. 4. and Reynolds numbers effects are small for turbulent condition.4. Several data points were eliminated based on a poor fit to the flight trajectory.4.5 Uncertainty Due to Boundarylayer Transition. Since the small difference between (1) and the current results can be eliminated by modifying the empirical constant appearing in (1). Because the error associated with fitting the theoretical trajectories to the experimental trajectories was not documented.3 Base Drag. Base drag taken from calculations made on the complete configuration agree with the trends cited by Love.4.5 to 7.4. and provided in Table 8. Figure 29 shows base drag from computations at a = 00 compared to base drag from a = 5*.4 Effect of Fins on Base Drag. The error for the reevaluated trajectories is summarized in 67 . The influence of the fins on base drag has been documented by Love [53]. Figure 30 shows the base drag from the complete configuration (Cases SB1SB4) compared to the base drag from calculations without fins (Appendix A). and is confirmed with the current method.
05 ...15 0.. . 0.0.10 0.  . 4 M.10 0.Empkiim "". 0. Viscousdrag coefficient. Cd. .. Basedrag coefficient.35  . . for the sharpnosetip model in turbulent flow (a = 00) 0. .25 0.. for the HART missile (a = 00) 68 ..00 .30 *4 0 ' 025 Civ0. I . Cd.05 0..I.40  GLE A(equaivdwa BOlAslan comnpinuible) Cuweat (fih nd) 0..00 1 .    1 2 3 4 5 6 7 8 Figure 27.1 Curna (Axiequs: Appmndix A) Cdb 0. 5 I 6 7 8 Figure 28.20 0.I 2 3 .I.20 "EAGLE 0..
15  . . .1 . I . 5 6 7 8 Figure 30. ..10 0. . showing fin interference effects (a = 0°) 69 . Basedrag coefficient. . 0. . . Cd. . Cd. .20 q 0. 4 5 6 Figure 29.. I . Basedrag coefficient. . .0.CuQvnt (Axieqns: Appendi A) Onem (3D.00 .  =5* Ct Cd.05 0 . . I 1 2 3 4 M. I 2 3 M.qn) 0.25 0.20 0.10 0. I . showing effect of angle of attack 0. . . . I . .00 I .05 0. .
the boundary layers on the fins are not well resolved. Also. This edge may disturb the boundary layer into a turbulent state prematurely. By eliminating those data with a poor trajectory fit. The computations with the fine grid (Cases S33S34) accurately resolve the boundary layer on the missile body. the highentropy layer of a blunt nosetip can delay transition [79]. giving much better results. This approach yields friction drag that agrees well with the flatplate theory of van Driest (van Driest II) [101]. The drag from the finegrid turbulent computations in Figure 32a is still lower than the experimental drag because of the coarseness of the grid in the azimuthal direction. turbulent bursts. but the computations using the coarse grid still underpredict the drag significantly. Second. induced transition may occur due to a lack of model fidelity. especially for the turbulent cases. In addition. the laminar drag trend with Mach number does not agree well with the experimental trend. fin friction drag is underpredicted. Several factors related to the nature of freeflight testing are important.6 (Cases $11$15 and S23S27. the turbulent fin friction drag used to obtain the total drag in Figure 32b is determined using (2) with a compressibility correction obtained graphically from White [101]. First. It is. Finally. respectively) are compared to the reevaluated experimental data in Figure 32a. Total drag from laminar and turbulent calculations at a = 0* for the sharpnosetip model and for Moo = 2 . Both coarsegrid and finegrid results are shown. can affect tranisition [42]. When the coarse grid results are corrected using this equivalent flat plate method.4. which are observed experimentally through shadowgraphs. The laminar drag severely underpredicts the experimental drag due to the much lower friction drag. The experimental missile body is a twopiece construction. natural transition has a Mach number and Reynolds number dependence [101. excellent correlation is obtained between the 70 . Subsequently. The turbulent drag predicts the trend with Mach number much better. the scatter is decreased considerably. Based on the excellent agreement obtained between computations of forebody friction drag and an equivalent flat plate (compressible).Table 9. however. and weak shocks can be seen emanating from the location where the two pieces are joined [93].6 Total Drag. 4. Due to inadequate resolution near the fin surfaces. still possible that the variation with Mach number is influenced by boundarylayer transition.
88 27.943 21.465  89090794 3.881 25.865 12.572 3.541 0.415 0..° 0.306 Roll error (deg)] 10.863 16.93 27.09 4.518 0.446 45.127 0.0010 0.473 0.0012 0.93  3.916 * m oC .563 Error between theoretical trajectory and experimental trajectory (from reevaluation of freeflight data) Lateral or Swerve error(m 0..200 10.577 0. 25.490 3.0021 Not use in this study due to larger pitch error.308 10.557 26.0019 0.462 23. (Yymm~d (original) 90012359 2.525 0.0017 0.128 0.45  3.670 Freeflight No.816 28.805 0.Table 8.0018 0.06 0.174 0.188 3.508 0.138 0..0016 0.0028 0.621 3.143 4.51 25.823 4. Experimental drag and pitchingmoment coefficients for sharpnosetip model with clippeddelta fins (a = 00) Free.556 * 3.370 5.194 4.0022 0.82 4.419 28.460 0.363 3.0012 0.678 * 89103032 90041706 90041203 90041304 90041605 Poor fit between Table 9.0021 Pitch Yaw error or (deg) 0.151 0.524 0..0014 0.677 3.598 9.725 3. 1 .239 0.0018 0.0023 0.119 0.0023 0.190 3.87 I I ____I (reevaluated) 2.132 0.452 0.416 0.flight No.0017 0.535  89090896 89090895 89053149 90011654 3.482 28. Longitudinal (yymmdd##) error (m) 90012359 90011957 89090896 89090895 89053149 89060150* 90020564* 89060758 90041706 90041203 90041304 90041605  0.255 0.67 24.206 0.123 4.913 89060150 89103134 90020564 89060758 90020263 90011856 3. 71 .443* 0.240 theoretical trajectory and experimental 0.0025 0.574 42.0017 0.588* 0.415 trajectory.554 42.801 0.418 30.264 7.0007 0.512 0.0022 0.C° ____ M C.0023 0.444 0.199 0.51 22.444 0.416 27.633 3. 29.0051 0.499 3.190 0.553 0.C (original)[ .42 4.710 24.461 0.24 90011957 2.810 6.363 3.230 0.445 0.418 29.461 0.220 6.982 9.0057 0.603 3.650 28.484 29.193 0.25 4.914 0.0023 0.811 2.758 19.267 50.501 * 28.410 0.717 27.
so extensive comparison is not possible. corroborate and/or identify the cause of the restabilization trend observed in freeflight testing.5. The pitchingmoment coefficient from the reevaluation of the freeflight data confirmed the change in stability trend. However. Laminar and turbulent calculations are made for only two Mach numbers. The secondorder polynomial curve fit that was used in Figure 34 is included for comparison. [37].6 and a = 0* is compared to experimental data (Table 10) in Figure 33. 4. The second goal of the analysis in this section is to determine the stability characteristics above Mach 4.5 to Mach 4. The static stability apparently increases (Cn becomes more negative) from Mach 3. it is apparent that the computations predict higher drag. Drag on the bluntnosetip model for Moo = 2 . Figure 34 shows the pitchingmoment coefficient obtained from the original experimental data for the sharpnosetip configuration (Table 8). Nearly identical results are obtained from the finegrid calculations when only the fin friction drag is corrected using the flatplate method. where experimentally determined Cn0 is plotted versus M". This trend is shown in Figure 34. First. the original experimental data for the bluntnosetip configuration (Table 10). as seen in Figure 32b.3.5 (Figure 4). As discussed in Section 4.5.4. 72 . There are two goals to this phase of the analysis. The data used in Figure 4 is summarized in Table 12. so accuracy for those freeflight tests is uncertain. both laminar and turbulent predictions exceed the freeflight data.experimental data and the present turbulent results. and recent experimental data for a grooved version of the sharpnosetip model (Table 11). Recall that for the baseline model. the static stability is assessed for the clippeddeltafin HART missile. The agreement between the curve fit and the reevaluated data suggests that the restabilization trend is confirmed with the reevaluation. the ZEUS and EAGLE computations differed markedly above Mach 4. a reevaluation of the raw experimental data was initiated by the author because the error for individual freeflight tests is not available. It is important to note that the bluntnosetip experimental data was not reevaluated. The pitchingmoment coefficient from the reevaluated data is shown in Figure 35 versus Mach number. A secondorder polynomial is fit through the combined data.5 Staticstability Analysis Next.
Cd.544 C.40 0.514 0. _ 72.60 0 0• 0 0 0 0.72 31.336 4. (yymmdd##) MC.531 0. extracted from freeflight tests of the sharpnosetip.731 0.0.593 0.653 1.125 4.01 27. HART missile Table 10. clippeddeltafin.640 0.30 2 3 i 4 5 Figure 31.393 I Cd I 91041741 91041740 91041742 91041843 91041944 91050248 91050147 91050249 [_d __M_ 0..86 27.369 4.64 42. Totaldrag coefficient.669 3.667 1.756 0.70 0.38 7171 71. 1. Experimental drag and pitchingmoment coefficients for bluntnosetip model with clippeddelta fins Freeflight No.708 0.070 3.22 73 .67 36.
80 "b 0 Experimenmt (eevaWNW~) Cnm n (tnturb ot fiat grgnd) 0 0.30  0. =W) o oCurrent 0. between computation and experiment for the sharpnosetip. . I & i 2 3 M. fine grid) 0.50 0 0. fl . Cd.und .30  *0. 4 5 6 7 Figure 32. .20 .. (b) fiatplate model of fin friction drag 74 . clippeddeltafin.I o ibd fa ictioa . HART missile (a = 00): (a) fin friction drag not accounted for. Comparison of the totaldrag coefficient. .40 0.20 r Cmece Ia.. 4 5 6 7 0.80 (F) Experiment (m .50 0400 0.70  Osminv.00.0.60 C 1 0. . coarm gri) 0Cunt Uuo. * ucia&ngd. I I 1 2 3 M.sviuaed) Cuniant (Imins.60 0 0 0 .
.0.80 0 0 Curme (Qaminr) 0.357 3.777 3.767 3.90 0.567 26.40 I I i .70 C4 0 0"' 0. C.. from freeflight tests of the grooved. 1 2 3 M. HART missile 00) °a= Table 11. 4 5 6 7 Figure 33. clippeddeltafin. Comparison of the totaldrag coefficient.794 24.50 0.958 3. (yymm##) 930710 930707 930708 930705 930709 930706 Moo C"_ 33.838 75 ..513 27..910 2. between computation and experiment for the bluntnosetip. Cd. sharpnosetip model with clippeddelta fins Freeflight No.260 3.60 0 0 0 0 0.178 21.226 24. Pitchingmoment coefficient.
5 5.0 8.0 7.41 17.0 3.0 EAGLE] 100. from ZEUS and EAGLE results for sharpnosetip model with delta fins Mw 1. &mooth body 2ndorder polyncuma ctnve fat.0 14.5 2.06 4.37 47.Table 12.26 59.0 17.05 89.2 1.0 4. Cm.0 ZEUS  43.37 12.0 12.0 6.7 12.42 110 100 90  Sai noseti... Pitchingmoment coefficient.69 0. smohbd so 70 0 Blunt nasezp. C. ail data C50 40 60• 40 0 30 20 10 0 1 I I 4 0 1 1 2 3 4 5 Figure 34. Pitchingmoment coefficient.0 23.5 3.0 32.02 2. for HART missile showing restabilization trend at higher Mach numbers 76 .02  25.5 4..
including the clippeddeltafin models discussed throughout this document. and still photographs of the models taken prior to launch. this is the fin size used in the postprocessing of the time. The records regarding the date and Mach number for individual tests show that 15 models were fired between 29 March 1990 and 17 April 1990. The photographs are digitized for inclusion in this thesis (Figure 37 and Figure 38). The pitchingmoment coefficient for all 45 tests is summarized by Figure 39. only 28 data points are shown.88). and several inverteddeltafin models.88 were fabricated later than the models for the 30 other tests. Based on shadowgraph photography from the freeflight tests. a line is used to distinguish the highest Mach number tests. Freeflight tests in the ARF (see Section 1. the test range log.1 for a description) are destructive. these 15 models correspond to the 15 highest Mach numbers (due to model failures Figure 39 shows only 12 data points above Mach 3. Also. therefore. Three key pieces of documentation are available.1. Several shadowgraphs appear to show different size fins on identical missile bodies [93]. More importantly. and orientation data to extract the aerodynamic coefficients [37].5.4. The construction of the models for these tests was not done at one time. Conversely. the author proposes that model inconsistencies might be the cause of the apparent restabilization. position. This size provides a surface area equal to the delta fin on the baseline model [93]. During the 19891990 time frame. Due to model failure. the construction and test logs indicate that the 15 models for the tests above Mach 3.1 Experimental Model Uncertainty. Figure 37 shows a model with a clippeddelta fin that agrees with the documented fin size and fin location. The purpose of Figure 39 is to propose a relation between when a model was constructed and at what Mach number it was fired. While it is not possible to associate specific models with specific tests. These are the model construction log. obtaining verification of model anomalies is difficult. Figure 38 shows that models were constructed with larger fins. The construction log from the Eglin AFB model shop indicates that 30 models were fabricated in May 1988 and 15 models were fabricated in February 1990. however. In Figure 39. and poor trajectory fits. three numerical experiments 77 . To further confirm this possibility. It is impossible to associate specific models with specific data points. The documented size is depicted in Figure 36. 45 individual freeflight tests were performed (sharpnosetip model). The author suggests that the higher Mach number tests were flown with models that had larger fins.
for HART missile from reevaluated freeflight data 78 . Pitchingmoment coefficient...evaluad data 2ndoder polynomial cme f all daa CUB 50 40 60 30 20 10 1 2 3 4 5 Figure 35.110 100 90 80 70 " Shap noset•p.. C..
all data from bluntnosetip experiments are used for comparison. Results from all thinfin calculations are summarized in Tables 1316. Largefin computations are made for Moo = 2.6 16 Documented fin size and position of clippeddelta fin (normalized by the body diameter) are conducted with a larger fin. The normalforce coefficient and the center of 79 .. respectively. The size and position of this large fin are determined solely from the photograph shown in Figure 38: the larger fin's root cord is 3. S35. and Mo = 4. These are designated Cases SB9.5.o = 3. = 3. In Table 13. and the distance from the missile base is unchanged..9. the results for the original fin size will be discussed at length in Sections 4. Spline curves are fit through the computed data in Figure 40. and S36. laminar. The smallfin data correlates best with the freeflight data below Mach 3. the information available suggests that the restabilization trend is the result of a model discrepancy and not aerodynamics. 0. and turbulent conditions.5..0 I Figure 36.5. Throughout the remaining analysis of static stability for the HART missile. The models for the bluntnosetip tests were constructed in February 1991. laminar.9 is treated as unreliable.24. the tip cord is 1.5. and S36 in Figure 40. and turbulent conditions. S35.6D.9 and the largefin data correlates best with the data above Mach 3.5. = 2 to Moo = 6 under inviscid. C. Mo. experimental data for any sharpnosetip test that exceeded M. C.8÷08 1. The stability analysis is now continued to determine the stability characteristics at Mach numbers exceeding 4.1.2D.6 0. and the fins for those models are believed to be correct. Although it is not possible to link the photograph of the larger fin to specific tests. = 2 to Moo = 6 under inviscid.4. is tabulated for the bluntnosetip computations for M. Similarly.. Therefore.. The pitchingmoment coefficient from computations with the original fin size (Cases $28S32) are compared to Cases SB9. in Table 14. is tabulated for the sharpnosetip computations for Mo.
A model that was constructed in agreement with the documented fin size and position Figure 38. A model that was constructed with a larger fin 80 .Figure 37.
I . 2 80 3 4 5 70 Sharp awes*p.60 (a) 70 MP so o CP Mfn Cno 20 10  . Experimental pitchingmoment coefficient.88 0 . inwnmdlt fins 50 C. (b) inverteddeltafin models 81 .: (a) clippeddeltafin models. 4 5 Figure 39. C. 3.88 I I 0 2 3 M. 40 30 20  10 I M"d.3...
5% of length to establish consistency with the experimental clippeddeltafin model. The interesting disagreements in Figures 41a and b exist above Mach 4. The locations of the centroid for the fins and the center of gravity for the clippeddeltafin model differ slightly from the baseline model. the magnitude of Cm. The coefficients in (10) are recalculated for a center of gravity located at 43. and the current inviscid results (Cases $1$10) in Figure 41a for the sharpnosetip model. with Moo at lower Mach numbers. The author believes the difference is related to the azimuthal clustering of nodes in the present study versus coarse unclustered nodes in the EAGLE study. In contrast to the EAGLE results. from Cases SISO agrees reasonably well with the EAGLE results. it also appears to exhibit a more modest decline in C. = 2 to A1o = 6. The data in these tables are also for M. Comparisons are also made to the experimental data to confirm the trends of Cm.5. and are discussed in Section 4.9) is equally well predicted by the current computations and previous EAGLE solver computations... predicted by inviscid. Third. Equation (9) is shown with the EAGLE results. the qualitative relation between Cm..2 Inviscid Results. linear theory. Second. each set of data indicate the decreasing trend in static stability with increasing Mach number. respectively.5. First.6. Sharpnosetip experimental data above Mach 3..pressure are tabulated in Tables 15 and 16 for the sharpnosetip and bluntnosetip computations. and Aio that is exhibited by the experimental data (for Moo < 3. Although freeflight data is apparently unreliable for Moo > 3. the current results show a much gentler decrease in the stability margin as Mach number increases beyond 4.88. Before boundarylayer effects (laminar and turbulent) are discussed. below AIoo = 4. The difference between the values predicted by the EAGLE solver and those predicted by the current solver are very important. The experimental data in Figure 41b are from the bluntnosetip tests and the reevaluated sharpnosetip tests. 4.9 is not shown in either figure based on the possible experimental model variations noted in Section 4..1. Neutral stability is 82 . the experimental data.5. the primary pitching characteristics are determined from the inviscid data in Tables 15 and 16. Several levels of agreement exist in Figures 41a and b. The inviscid pitchingmoment results are primarily compared with EAGLE results and the theory provided by (9) and (10).
58. 4 5 6 7 Figure 40.0 Table 14.5 6.601 70 60x so 50 40 30 'a . .93 19. Pitchingmoment coefficients from the correct fin and the larger fin Table 13.0 66. Pitchingmoment coefficient.00 29.46 23.73 34.0 4.5 4.0 3.89 42.37 31.64 38.0 3..32 35.20 14.36 33. 20 10 size (Cow S2S32) $fin I I I I 0 I 1 2 3 M. C.79 27.34 26.60 27.83 28.34 10..64 17.14 24.44 Pitchingmoment coefficient.69 27.50 57.26 11.54 83 .03 23.5 4. from current computations for sharpnosetip model with clippeddelta fins (Cases SIS34) MMol Inviscid Laminar Turbulent 2.06 24.51 22.5 5.0 2.. C.63 30.0 6. from current computations for bluntnosetip model with clippeddelta fins (Cases BIB31) Moollnviscid 2..17 20.59 30.79 27..73 15.72 50.53 34.89 21.00 43.96 Turbulent 58.95 3.46 Laminar 64..
0 3. C.09 Table 16. from current computations for bluntnosetip model with clippeddelta fins Moo 2.01 0..538 0.0 3.48 0.582 0.62 7.50 8.642 0.92 10.564 0.556 0..508 0.83 8.519 0. Turbulent xc..56 8..23 9.05 7.577 0.557 84 .0 6.504 11.5 6. and center of pressure.631 0.489 0..81 8. Normalforce coefficient. xqI from current computations for sharpnosetip model with clippeddelta fins M00 2.° J Laminar XplL Turbulent C..0 Laminar C.03 7.88 10..5 5.611 0.621 0.89 10.08 9.plL Xo Turbulent C.84 Laminar z. Normalforce coefficient.15 9.549 0.o Turbulent xmlL 13.50 9..5 4. and center of pressure.0 3.569 12.570 0. xz.0 Laminar 12. C..Table 15.564 0. lL 0.
tip.. smooth body body E. inviwd 6....5  S..'m 40 .. 50 '.. for clippeddeltafin model: inviscid results: (a) sharpnosetip computations versus theory.. . (b) Exient sharp n... Pitchingmoment coefficient.. ..90 0 . " . "C. ip.. 20 .a.Expiment SIt..omhip. (a) :bop nowtip.. invisid 50 40 C. (b) sharpnosetip and bluntnosetip computations versus experiment 85 .. 20  10  N~mal Stabili• 0 10I 1 2 3 4 5 6 7 8 M_ Figure 41.. xwed EAGLE Iiner theory: Equation (14) Currnt: sharpn. experiment.. smooth body EwIIL blunt amc* smooth body Ezpesznunt sarop noselip.45 30 10 210 Net Stability . • .... .. groovod body Cunt sharp nursip.. invisd Currmt: blunt mip. menL sbwp mc.. 10 I I I 1 2 3 4 5 6 7 8 9080 o 70 ... C. and EAGLE results.
86 . 20 10 0 . Expmainu: sbip d. the present computations depict a trend with Mach number that is between the EAGLE trend and the ZEUS trend. wmoot body Eime: srp ncu•p. C.. at Moo = 2.. .5... Above Mach 4. Comparison of pitchingmoment coefficient... the difference is believed to be related to azimuthal clustering near the fin surfaces.. and beyond that with the bluntnosetip results. C.. Neutal Stabiliy * 10 I I 1 2 3 4 M... and persists for the laminar and turbulent calculations.so o 70 60 . as shown in Sections 4. invisd C 40 30 .. the Cn. from three inviscid solvers reached at Mo....5.m with MO.•Mo trends for all inviscid computations (prior and present) on the sharpnosetip model are compared. At M.4. Cm for the blunt model is twice that for the sharp model. below Mach 4. Again.. The experimental data are included for reference. In Figure 42. Vroved body EAGLE ZEUS m =m:sharp osetip.34. All three sets of computational data predict similar trends in C. at higher Mach numbers. This is directly related to the nosetip geometries. In Figure 41b... . due to a shift in the center of pressure on the missile body. = 7 with EAGLE. but beyond Moo = 7 with extrapolation of the current sharpnosetip results. The ZEUS results from the baseline deltafin configuration are included to provide useful qualitative information at higher Mach numbers. = 6.. 5 6 7 _ _I 8 Figure 42. The bluntnosetip results differ from the sharpnosetip results. for the bluntnosetip model is 13 % more negative than the sharpnosetip model.
The effects of viscosity are initially assessed for laminar boundarylayer conditions. but the impact of turbulence is best determined after laminar results are documented. In Figure 45. a = 5 0. and Moo = 2. the boundary layer appears to improve C. The crossplane pressure contours in Figures 4655. z = 0). however.25. is more negative). This is evidenced by the piling up of pressure contours on the bottom of the leeward fin. C. the velocity vectors show that the crossflow is small. the pressure gradient near the fin surfaces increases.5. At higher Mach numbers. also indicate that the laminar flow structure is similar to the inviscid structure below Mach 3. the boundary layer has influenced the pressure distribution 87 . Most experiments indicated that the boundary layer had become turbulent upstream of the fins. As Mo. the laminar and inviscid results agree at low Mach numbers. Interestingly. In addition. the sharpnosetip and bluntnosetip pitchingmoment coefficients from the present study are plotted versus Mach number with the experimental data (Figure 44). and laminar flow (Cases S16S20) are compared to those for inviscid flow (Cases S6S1O) in Figures 4655. These crossplane contour plots are for E = 21.5. The shape of the streamline which begins on the windside of the missile near the symmetry plane (y = 1. The inviscid results that were previously discussed are included for reference. increases.e. It also shows that a refined grid is needed in this area (clustering near fin surfaces is employed in all computations in this study).3 Laminar Results. At the higher Mach numbers. Streamlines created from the crossplane velocities are also shown in Figure 43. = 2 to Mo. and the expansion and compression on each fin are distinctly separated.8. M". and begin to diverge at about Moo = 3.. The pressure contours in Figure 45 are very similar to those for inviscid flow (Figures 22 and 23).5..o (i. The pressure contours in the cross plane (yz plane) for ca = 5°. D This is a point near the middle of the fins. The agreement at low Mach number is also evidenced by the fins' surface pressures. = 6. 4. shows that the cross flow is not significantly displaced from the missile surface (expect for the region above the leeside fin).. The impact of the boundary layer is approximately the same for the sharp and blunt models. the pressure contours are shown on the fin surfaces for laminar flow. Again.The crossplane velocity components are shown in Figure 43 for inviscid flow at Moo = 2 and a = 50 (Case S6).
25 t I III 'i 1.25 0.1.00 Figure 43.00 0.25 t 7 / 111 I 0.50 1. = 2..50 0.75 .00 z .50 2.00 0.25 0.' f! I "•I I! I 1.50 0. Crossplane velocity components.75 \\~i I 'kk 'i 0. c = 50) 88 . 1 1. Case S6 (Mo.00 '1 0.
. C. mviscid CIuret: Sharp nose04. smooth body Expeait: sharp noeip. mvisi CurenMt blnt nogetip.90 so 70 h Eimut: sharp noseip.. laminar 50. Expaimot blut noetip.. for clippeddeltafin model: laminar results 89 . 040 k... v\  Current sharp nosetfip. lamina Crrnmt blut nosecip. grooved body . 30 200 10 0 10 Nui'ral Stability 1 2 3 I 4 1 5 I 6 I 7 I 8 Figure 44... Pitchingmoment coefficient.. smooth body S 40o ....
80 1. (c) upper surface of the fin on the compression side of the missile.759 1.2145 0.2265 02148 02M4 r 115 1.00 8 7 6 5 4 (d) 0.5 330 X 6 4 0.234 0.L.0 21.5 .0 M M 3.1571 0.75 1.3034 0. 233. P.00(d 1.M4 0.0 X 215 23.1.5 1.0 2.5 33.00 A 1. M3. (b) lower surface of the fin on the expansion side of the missile.50 A 9 a 7 (a) 0. (d) lower surface of the fin on the compression side of the missile 90 .2331 0.326 0.0 21.75 21.1571 2.25 1.5 Figure 45.1681 4 3 2 3. 3 0.1 0.0. Pressure contours on the fin surfaces for Case S16 (M.80 r 1.2149 0.5 22.0 as5 \3 23.25 1..80. = 2.0 21.1667 ~~"~~ J 0. or6~ r 1.00 A 9 9 ? OX.0 01667 0.0 23.1911 0.60 A 9 a 7 (b) OM.31 21.25 3.2265 02149 0.1918 r. O.uss 0.1002 0.1916 0.5 M 3.•3. 0.0 31.ao 1.5 2Loo 1.00' 6M 5 01.000.75 1ow 0.75 1. =5): (a) upper surface of the fin on the expansion side of the missile.75 1.1455 3.5 33.333 0.
75 1.1802 0.50 025 8 7  0.2380 0. Pressure contours.75 4 A 9 0.2264 8 7 6 0.1571 0.50 0.1918 5 4 3 0. laminar) 91 .2264 0.125 1.25 0.50 0. Case S16 (Mo.1802 0. a = 50. a = 50.1340 S1 2 z 225 Figure 47.50 o751 2 0.00 0. Case S6 (Moo = 2.1571 0.00 0. inviscid) 1251. Pressure contours.75 0.50 025 y 0.00 0.2149 0. I 0.1918 0.2033 0.1687 0.2380 0. = 2.50 6 5 4 3 0.75 . 1.001 2 5 " 0.00 025 A 9 0.75 1.1687 3' 0.00 0.2149 0.00 0.1456 0.751 1.2033 0.1340 AM 225 Figure 46.1456 0. z .00 .
Pressure contours.0801 0.0852 y 0.000.0955 0. Pressure contours. inviscid) 1.00 055 6 4 0.25 ýI.25 0. 0.50 025 A 9 0.0904 0.75 0.00 .1058 0. = 3.0646 0. 3 2 0.25 1. Case S17 (Moo = 3.25 A 9 8 7 4 0.0852 y 0.0646 0.75 z 1.1007 8 7 6 0.1058 0. 0. .50 225 z Figure 49.0696 0.0801 0.00 1.50 0.0595 .251.00 025 5 4 7 0.00 1.0698 3 0. laminar) 92 .0955 0.75 0.0749 0.75S1 1.0595 I 1.0749 0. 0.0904 0.50 0.75 2 1 .50 225 Figure 48.00 1.00 0. a = 50.750. a = 5o.1.1007 0.50 0. Case S29 (M.
0.0590 0.50 2.75 z 1.5.125 1.50 0. = 3.00 0.00 0.00 0. Case S18 (Mo = 3.75 0.0440 225 Figure 50. inviscid) 125 1. Pressure contours.75 A 9 8 7 6 5 4 3 0.0665 0.0477 0.0627 0.0552 0.0627 0.50 0.75 z 1.0440 125 0.25 Figure 51.0477 0. a = 5°.00 125 .oo0 0254 0.00 0. Case S30 (M.25 A 9 8 7 6 5 4 3 0.0250.0740 0.00 0.0552 0.0702 0. laminar) 93 .0702 0.00 .0o7 0.0515 0.0590 0. a = 5°.50 S2 1 1.5• 2 1 1.0740 0. Pressure contours.0777 0.5o 025 y o.0665 0.5.0515 y 0.
50 I 225 I Figure 52. Pressure contours.0356 0.00 025 0.0470 0.75 0.00 6 025 5 4 0.5.00125' 3 0.125 0.751.00 0. a = 50.00 0.752 1 0.50 0.0447 0.00 0.0265 0.0402 y 0.25 9 8 7 4 0.50 0.75 0.0424 0. Case S31 (M.125 1.0402 0.00 0.75 z 1.0379 0.0424 0.50 0.75 A 0.0288 0. inviscid) 125 1.50 225 50. Figure 53.5.75 z 1.0265 1.0447 8 7 0.0470 0. :6 14 41'3 0.0288 ' 0. = 4. Case S19 (Mo = 4. Pressure contours.00o .0311 2 1 0.50 025  4 A 9 0.0311 0.033 0.0333 0. a = laminar) 94 .
0162 0.00 .00 0. laminar) 95 . = 6. 75 z 1.• 0254 0.00 0.72 1' 0.50.0252 0.0175 0. a = 5°. Case S20 (Mo.0149 120.01725 0.50 225 Figure 55.00 025 AM 0.50 025 0. 0.75 / (4 6 4 3 2 1 0.50 025 y 0.0187 0.50 225 Figure 54.00 0.2952 0.0149 125 '.0239 0.0226 0.75 A 99 0265 0. Pressure contours.7 1. a = 5*.125 1.0213 0.75 z 1.0187 • k 2 3 1 •62 1 0.75 A 9 8 7 1 0.0. 0. Case S10 (Moo = 6.0265 0. Pressure contours. inviscid) 1257 1.00 0.
The influence of turbulence is now addressed. The pitchingmoment coefficient for turbulent flow increases measurably over that for laminar flow. The sharpnosetip and bluntnosetip turbulent results are plotted with the laminar results in Figures 57a and 57b.4 different transition points (Section 3. respectively. less negative) for the turbulent computations is due to the compression bleeding around the leading edge of the windside fin. The bleeding of higher pressure air around the fin leading edges is visible in the turbulent case. The crossplane pressure contours for turbulent flow. and Mo.slightly. The lower static stability (C.. This indicates that the turbulence plays a significant role in the region where the cross flow is blocked. Turbulent Results. The streamlines are significantly displaced from the missile body between the windside fin and the symmetry plane. The bluntnosetip re96 . Figure 56 shows the crossplane velocity components for laminar flow at Moo = 2 and a = 5* (Case S16).. Also. the same modest trend with Mach number is seen for both nosetips at higher Mach number. the pitchingmoment coefficient decreases less with increased Mach number owing to the presence of viscosity. a = 50.. The bleeding effect for turbulent flow is also seen in the crossplane pressure contours. The crossplane velocity components show that crossflow is small for laminar flcw also. and shown in Figure 56. Recall that the sharpnosetip models and the bluntnosetip models have 4.. = 6 can be seen in Figure 55 versus Figure 54. Streamlines are again created from the crossplane velocities. = 2 (Cases S28) are compared to those from laminar flow in Figures 59 and 60. These streamlines are slightly more displaced than those for inviscid flow. The position of the transition point does not seem to have a large impact on C. The cause of this phenomenon is seen in the streamlines of the crossplane velocity components (Figure 58). is within about 10%. since sharpnosetip results (Figure 57a) and bluntnosetip results (Figure 57b) show similar decreases in Cn.4). The experimental values are added for comparison. and affecting the pressure on the expansion side of the fin. Thus.5.. The higher pressure values on the compression sides of the fins for M. from the laminar results. Figure 61 compares the turbulent pitchingmoment results for both nosetips. Agreement between the computed and experimental values of Cm.
50 1. Crossplane velocity components.25 1. Case S16 (Moo = 2.50 2.00 0.25  1 _I 1.25 I lIf 0.00 z .00 Figure 56.50 0.50 .1t .75 1 0.00 I 1._ 1. Ct = 50) 97 .1.__ 0._ tt It 1.00' 0.25 0. i.
*55 30 . C 0 . uk N: Mug f.90 (a) 70  . Curre: bubitna.. for clippeddeltafin model.... mn: alnpna ip.." Nwlm SUabilift 1 2 3 4 I LI 5 I 6 7 I 8 Aof) (b) 70 60 40 ..UnbulM 90 C. turbulent results versus laminar results: (a) sharp nosetip. UiamUW . Pitchingmoment coefficient.... (b) blunt nosetip 98 . 40 30 20 10 0 10I "... .• .. 101" "4.. . .I up.. .. 20 1o 0 10 Ne n OWa Stabili 1 I I I p I 2 3 4 5 6 7 8 Figure 57.. . Cam: sp pifhdp.
J t 0.00 z 1.75 0.25 1.00 .25I 0.50 2.25 .25 t 0. a = 50) components. Case S28 (M Figure 58.50 0.50 0.1.00 0.50 1. Crossplane velocity 99 .75 . 7.00 0.= 2. 1.
00 . a = 5'.00 0.1687 y 0.2264 0.1571 0.1456 0.2149 0.00" 0.1.75 z 1.1340 225 Figure 60.125 1.71 4"•A 0.1918 0.00 0.2380 0. laminar) 100 .75 o..00 0. Case S16 (Moo = 2.2149 0.751. 250.1918 0.75 1. a = 50.251 0. Pressure contours.2264 0.1456 0.25 Figure 59.50 0. Case S28 (Mo = 2.1802 0.2033 0.1571 0. 3 2 1 0.2380 M 0.25• 9 8 7 6 5 4 0.1802 0.00.50 / 0.1340 0.50 0.50 2. turbulent) 1.00 1.00 . .75 1.50 025 A 9 8 7 6 5 4 0.1687 y 0.2033 0.25 . Pressure contours.25 1. 5 0. .50 z 3 2 1 0.75 0.
1 Change in Centerofgravity and Fin Position. These computations are made for a = 50 and Mo = 2 to M. 4. while the clippeddeltafin model has a center of gravity at 43.6 Comparison Between the Clippeddeltafin Configuration and the Baseline Configuration. the locations of the fin leading edges differed.5. characteristics for the clippeddeltafin model are contrasted with the traits of the 101 . Two differences that do exist. The deltafin model has a center of gravity at 43% of length. the base flow did not appreciably affect the pitchingmoment coefficient (Table 13). In order to com pare the results for the clippeddeltafin model to the baseline model.5% of length. which has delta fins.5. Given the differences stated above. Figure 63 shows both fins and their positions on the missile body.5 Influence of Base Flow. 4. Throughout the range of Mach numbers tested.5. similarities and differences in the models must be taken into account.6.2 Aerodynamic Comparison.suits provide further clarification to the suspected restabilization phenomenon. and thus better stability. The different leadingedge locations will also affect the pitchingmoment coefficient..5. Both models had identical tangentogive noses with smooth cylindrical bodies. The variation of pressure across the missile base is mild (Figure 62). In addition to a slightly different center of gravity. values than the sharpdata curve. and equivalent fin surface areas. the bluntdata curve has more negative Cm. 4. The aerodynamic coefficients (drag and pitching moment) are then contrasted to determine the impact of changing the planform of the fins. A more favorable pressure distribution on the missile body causes a small shift in the center of pressure. Four computations are performed to assess the impact of the base region on the pitchingmoment coefficient. First. differences between the two configurations are discussed. The present computations indicate that the stability of the blunted model is superior. The correlation of the bluntnosetip experimental pitchingmoment coefficients with the erroneous sharpnosetip pitchingmoment coefficients was purely coincidental. 4. Although this may not seem very dramatic. = 6 (Cases SB5SB8). The effects of the base flow on pitching moment are examined next. thus producing almost no contribution to the pitching moment. it does affect the pitchingmoment coefficient measurably. however. are the centerofgravity locations and the finleadingedge locations.6.
0 . smooth body . . . 2 0 c ll a l b ...t sup no"tip..90  so a Expaiinat sharp nos*i. . 40 0 " If.. ububt blun nosetip turbulmt 50 C.. . 10 0 10 Neutral Stability I I I I I 1 2 3 4 M. Pitchingmoment coefficient. C. . . 5 6 7 8 Figure 61. smooth body 0 70a Evaiernit bl=n nobep.. ... . rooved body CuuM.Curzent 7 Expdmeat sAmpnowe*i. for clippeddeltafin model: turbulent results versus experiment 102 .  30 20 *1 A 5 •..
0257 0.0242 0. The effect on drag is minor.75 Figure 62.Moo curve for the clippeddeltafin configuration is slightly less steep than the curve for the deltafin configuration. The difference is smaller at higher Mach numbers. and compares it to the current turbulent results. despite the small discrepancies in the models. The wave drag on the fins of the deltafin model should be slightly greater than the wave drag on the fins of the clippeddeltafin model. The direct comparison indicates that the clippeddeltafin has slightly inferior stability characteristics at lower Mach numbers.25 A 9 8 7 6 S5 4 3 2 1 0.0170 0. and the drag data from the clippeddeltafin freeflight analysis.0213 0. Figure 64 combines the drag data from deltafin freeflight analysis. the slope of the C.. The deltafin model demonstrates better (more negative) stability due to the more forward center of gravity and the more rearward fin position. Thus.25 0. a = 50) baseline.0199 0. This difference is impossible to detect within the scatter of the data.0228 0.50 0.0184 0. Case SB7 (Mo = 6.50 0. Pressure contours on the missile base.00 0. the experimental data is in agreement. The pitchingmoment coefficients from both models are shown in Figure 65. In addition.0.0271 0. The staticstability characteristics differ much more. deltafin model.50 0.0300 0. but potentially superior characteristics at higher Mach num103 .0286 0.
6 0. As stated in Section 4. at higher Mach numbers.1. = 6. 104 . for the present calculations. The author attributes the difference to greater portions of the fin area being submerged in the boundary layer. This is supported by the present viscous calculation which extend to M.5.261 1.6 Figure 63.8 082 2.3. Positions of the clippeddelta fin and the delta fin relative to the base bers.0 0. the boundary layer affects the pressure distribution in a manner which increases fin effectiveness.
comparison between deltafin missile and clippeddeltafin missile 110 100 70 0 aDelafin Cppdmds~a fin.. bwoduay C ° 0% .30') 1 2 3 4 5 6 Figure 64.. comparison between deltafin missile and clippeddeltafin missile 105 . Drag coefficient.60 0 OAO a uD o DfiuMe Clipped. Pitchingmoment coefficient.40 0. 5 6 7 8 Figure 65.. C.10 1 2 3 I I I I I 4 M. 0 00 0.70 0.0.50 0c.. 0 CmREbLdbirbulwt •' C 50 aonS 400 30 20 40 0 ý 0 . Cd. smooth body OzpedWM fh oin. skAim fligh) 0.50 0.
Using the present developed technique. as demonstrated by the finegrid calculations on the sharpnosetip model. An infinitely thinfin assumption is used to decrease the grid requirements. all forces on the body and normal forces on the fins are accurately modeled with the present method. with or without a base region. the flow structure between 106 . These overcome the deficiency in drag prediction from the thinfin assumption.. and the method can be used on sharpnosetip or bluntnosetip configurations. For the latter. The accuracy to which shock waves. Both drag and pitching moment are dominated by compressibility effects. The discrepancy between the present prediction of C. For the small angle of attack used in this study. Most experimentally determined pitchingmoment coefficients fall between the present laminar and turbulent calculations (Figure 66).25). Inviscid. around MO" = 2. The expansion and compression associated with the upper and lower surfaces of the inclined fins produce less net lift for the present turbulent computations than the present inviscid or laminar computations. For the former. the base drag and turbulent viscous drag decrease substantially with increasing Mach number. Due to the degenerate cross sections of the fins. or turbulent conditions can be simulated. laminar. forces on fin edges cannot be directly calculated.. an accurate upper limit on drag is established. the presence of turbulence is found to be significant. The base flow is found not to affect the pitching characteristics significantly. Summary and Conclusions from Thinfin Investigation An efficient and accurate computational tool has been constructed to predict the aerodynamics for finstabilized missiles in supersonic flow. a = 50.5. This limit corresponds to turbulent conditions over most of the body (V > 1. and the experimental prediction of C. boundarylayer properties and baseflow effects are calculated is documented over a broad range of Reynolds numbers and Mach numbers. By using an empirical correction for the viscous drag on the fins (required due to the coarseness of the grid in the azimuthal direction). fin effectiveness is seen to decrease with increasing Mach number.9 cannot be explained by the author.. so semiempirical corrections are proposed for fin wave drag and fin base drag. Also. In general. the flow about the complete Hypersonic Applied Research Technology (HART) missile configuration is simulated.
The clustering of nodes near the fin surfaces may explain the difference. are predicted. C.. I I I I 1 2 3 4 5 6 7 8 M... at Mach numbers beyond the experimental testing capabilities.40  A paim dpos•dp.. Qrm sbap netp. In addition. the variation of pressure across the base does not contribute measurably to the pitching moment. Figure 66. urbulmt It ''s. The current results agree more closely with the trends predicted by ZEUS. no$*. = 4 are not fully explained. The aerodynamic characteristics of the HART missile.. decreases with increasing Mach number much less than the previous EAGLE computations. The larger than expected drag scatter. = 7.. that is seen experimentally for the sharpnosetip configuration. The current predictions indicate that C. is explained. the fins and the base does not change significantly from the structure for simulations made without a wake.90 70 so 50 Cm. for sharpnosetip HART missile with clippeddelta fins (a = 0*). •' 0 . = 6 suggests that the clippeddeltafin configuration is stable past M... b smoolk body a . between the current inviscid results and EAGLE results above M. INp.  20 10 0 10I Neutrl Stabiliy. A reevaluation of the experimental data 107 .. Extrapolating the current results beyond M.. E . and should be explored further for future EAGLE computations. Pitchingmoment coefficient.unet uhap asehp. The differences in C.W wvedbody C.
This suggests that for most flights boundarylayer transition does occur on the ogive nose. but the contribution to pitching moment from the missile body is demonstrated to depend on nosetip shape. deltafin configuration reveals two important points. As expected. A direct comparison of the drag and pitching moment between the clippeddeltafin configuration and the baseline. Documentation is provided to show that a larger fin was used in some of the freeflight tests. Although the wave drag on the fins of the deltafin model should be slightly greater than the wave drag on the fins 108 . This trend is seen only in the experimental data. The poor correlation between previous computational predictions of drag (EAGLE with corrections) and experimental drag is also explained. a reevaluation of the data. Boundarylayer effects are the presumed cause for the remaining scatter. as modeled. computations were performed with a larger fin. and is believed to be the result of model discrepancies. and very close to the latter. particular as M. the effect on drag in minor. An explanation for the restabilization trend with increasing Mach number has been offered. using the correct smaller fin size. and for a few flights the boundary layer was laminar over a larger portion of the forebody.reveals that much of the scatter is due to poor fits between theoretical and experimental trajectories. The empirical relation used for turbulent viscous drag does not take into account compressibility and the large Mach number dependence exhibited by turbulent skin friction. and results in better staticstability characteristics. Fin effectiveness decreases due to compressibility by an amount approximately equal to that for the sharpnosetip model. The pressure distribution on the bluntnosetip missile body affects the center of pressure significantly. Finally. The reevaluated drag data falls between the current laminar and turbulent calculations. increases. confirmed the apparent change in the staticstability trend with increasing Mach number. drag is much higher for a blunt nosetip. and these calculations agree with the suspect experimental data. In addition. The same causes (neglecting compressibility effects and fin pressure drag) are traced to the mediocre agreement between corrected EAGLE results aud experimental results for the baseline HART missile (deltafin configuration). In this study. fin pressure drag is neglected. In addition. The poor agreement is caused by the empirical corrections not the EAGLE results. it is determined that the wave drag and base drag on the fins are not negligible. The effects of nosetip blunting are demonstrated. First.
the boundary layer affects the pressure distribution in a manner which increases fin effectiveness.= 6. the difference is within the scatter of data.. The deficiency in the pressuregradient boundary condition is overcome by clustering nodes near fin surfaces. a spacing of y+ t 1 is needed with at least 40 points (cells) in the boundary layer. The author attributes the effect to greater portions of the fins being submerged in the boundary layer.of the clippeddeltafin model. Therefore. boundarylayer length scale. In terms of the nondimensional. y+. It is demonstrated that accurate calculation of turbulent skin friction necessitates a very small node size near the missile surface. This is because the effects of the base region on pitching moment are small.. The grid requirements are found to be demanding to simulate flow for a complete missile configuration. and indicates that the clippeddeltafin configuration has slightly inferior stability characteristics at lower Mach numbers. The requirements are less to predict other boundarylayer properties and surface pressure with reasonable accuracy. Second. This remains an unresolved issue. the pitching characteristics of the HART missile are accurately predicted with the thinfin assumption. The slope of the C. This is supported by the current viscous calculations which extend to M. Simulating the flow in the base or wake region adds considerably to the computational requirements. First. a reasonably small number of nodes are required in the azimuthal direction (33 for the bisymmetric half plane in this study). Eliminating the wake region decreases the overall number of grid points required. at higher Mach numbers.M. 109 . it is not necessary to simulate the base flow in order to predict Cd or Co. the staticstability characteristics differ markedly. It is important to emphasize that these requirements are needed to achieve grid independence for skin friction. Two simplifications are found to greatly reduce the grid requirements in general.. and viscous calculations for the deltafin configuration are warranted to further discern the cause(s) of the differences. For the present computations. In addition. curve for the clippeddeltafin model is slightly less steep than the curve for the deltafin model. put potentially superior characteristics at higher Mach numbers. and the drag on the fins is calculated using the degenerate cross sections with semiempirical corrections.. and the accuracy of the empirical base drag relation is reasonable.
Part II: Investigation Using Thick Fin 110 .
and the crosssectional shape of a fin. is addressed in this part of the research. the amount of fin area submerged in the missile boundary layer. the effect of fin thickness may be tertiary. As stated above. In general. The precise values of properties within the boundary layer on the fin will impact the pitching moment at a tertiary level. 111 . these include viscous effects. by virtue of the fin's impermeable nature. It is possible that the shock waves created by the highly swept fins do not induce boundarylayer separation. Therefore. supersonic flow past a swept. shape. location. and thereby the effect of thickness and cross section. the flow structure near a bluntsweptfin/body junction is reviewed with the emphasis on highly swept fins. 6. First.6. and orientation of a fin will dominate its pitching capability. the size. and would depend on the average thickness of the fin in causing flow blockage. Unfortunately. Obviously. and is also discussed.1 Increased Modeling Complexity The intent of this part of the research is to determine whether the thickness and crosssectional shape of the fins influence the pitching moment of the HART missile significantly. to the best of the author's knowledge. flatfaced fin. In this case. This suggests that the impact of the fin's thickness and cross section may in fact be small. the numerical implementation of the thick fin is presented. The secondary viscous effects were examined in Part I by neglecting the impact of crosssectional shape and thickness. Other factors will affect pitching moment and static stability in a secondorder manner. This was seen in Part I with a series of inviscid computations. A boundary for incipient separation at the finleadingedge/body junction was suggested by Stollery [80]. The accuracy of this assumption. Finally. Methodology for Thickfin Investigation Many aspects of the fins influence the stability characteristics of a finstabilized missile. The results from the infinitely thin fin assumption in Part I agree very well with the experimental data. no quantitative information is readily available in the literature. the author has categorized the effect as secondary. nobody has investigated turbulent. not on the actual crosssectional shape.
swept fin at incidence are taken from [41] for A = 30* and A = 750 and M. To gauge the benefit of sweep in reducing the level and extent of the disturbed pressure field. = 2. To further clarify the thinfin results. The current turbulent thin fin results (A = 70°) are also shown in Figure 67 for Case S29 (M. Stollery [80] suggests that there is a direct relationship between the Mach number and sweep 112 . As discussed in Chapter 4.4. however. The higher pressure in the interaction region. but is from a line projected upstream from the junction. Several researchers have questioned whether separation occurs for severely swept. upstream of the finleadingedge/body junction. This was discussed in Secion 2.95 and a = 50). and the development of a horseshoe vortex structure [80]. there is a general lack of data for flatfaced fins. which influences the flow near the body and around the fin. 6. Both the pressure ahead of the fin/body junction and the pressure immediately downstream are affected significantly by sweep. = 2. near its root. 55]. Because the fin from [41] is spherically blunted. and the lower pressure associated with the vortical flow region may in turn modify the staticstability characteristics of the missile. calculations are made for two Mach numbers and a thick fin. The interaction weakens dramatically.95 (Figure 67). the boundary layer does not separate in the current computations with the thinfin assumption. t. cylindricallyblunted fins at high Mach numbers [64. The pressure is measured along the junction line and nondimensionalized by the pressure 5 fin thicknesses. for a blunt fin.2 Flow Structure Near a BluntSweptFin/Body Junction The flow structure which should be predicted for this fin geometry is not entirely clear.6. some data does not correspond to the fin/body junction. In general. Again. This results in flow reversal.2. a fin/body junction results in the formation of a corner vortex.3 Suggested Boundary for Incipient Separation at Fin Leading Edge The strength of the interaction at the fin/body junction relies on the sweep angle of the fin. P. Also. it is possible (definite for an unswept fin) that the shock created by the fin separates the boundary layer on the missile body. as the fin sweep angle increases. 99. the pressure distributions generated by a blunt. This could be due to the severe sweep angle. it may be possible to identify more precisely the sweep angle required to prevent separation.. However. as suggested by the data from [41].
5 . everywhere else. The first condition is the maximum sweep angle that ensures the leading edges are supersonic. angle such that a boundary for incipient separation exists..00 2. a 5 2.75 2. in Figure 69.Mo. Therefore. separation is possible.Buatfla/iaspAm (ROW.00 3 2 1 0 1 2 3 4 5 6 7 Figure 67. a line is added to show where incipient separation occurs in supersonic flow over a compression corner [80]. The boundary is established by two conditions required for an attached shock at the nose. region defined by this boundary. The region above this line is where an attached boundary layer is likely.00 1. in the sense that the normal component of the Mach number is greater than one (MN > 1). 113 . the line representing separation over a compression corner helps to further define the region where the shock is attached at the leading edge. 70): A 306. A70. Surface pressure from a sweptbluntfin/flatplate interaction and the current thinfin/missilebody interaction.75.75 1.3. Also. the two Mach numbers examined in Part II of this study are indicated. Kubota and Stollery [48] have shown that separation occurs more readily for a glancing interaction (fin/body) than for a twodimensional interaction. a5" 2252..50  Cunam (CmsS29): Cwfw (Cs S29): *135".A 45". it is not. and below the line it is only possible. In Figure 69. The region bounded by these two conditions is shown in Figure 68. Within a A . a5' 706.50 125 1. 70): A . aw5" Bimtfi wftaw (OW. The second condition is the minimum sweep angle required for an attached oblique shock to a wedge.
The shape of the inflow boundary is taken from grid points used in Part I. The point M. = 2 and A = 70° lies outside the "attached" region. and the point M. The domain size is reduced to add more cells in the finned region. the entire configuration is not reexamined.2. Boundary for an attached shock at the leading edge of a blunt fin. 6 7 8 I 9 10 i Figure 68. Instead. The thick fin is modeled using an Hgrid [86] to minimize algorithm and boundary condition changes. the outer boundary is placed only three body diameters from the missile surface. The new boundary conditions are described in Section 6. In addition. 6. allowing for potential disturbances ahead of the junction.4 Numerical Implementation Due to increased computational demand.. The structure of the grid and nodifications to the boundary conditions are discussed in the next two sections. 6. a much smaller region is defined around the fins and aft part of the missile.4. = 6 and A = 70° lies well within the region. 114 .90 70 601 A 40 30 20 10 0 I I I Minimum A for autthed oblku shock to a wede.4. Both of these boundaries require different boundary conditions than used in Part I.. necessitated by grid refinement near the fin leading edge. The inflow boundary is placed ten fin thicknesses upstream of the finleadingedge/missilebody junction.1 Grid Structure. 1 2 3 4 5 M.
. The deviation from a normal distribution is small... A total of 121 cells are distributed in the azimuthal half plane. and the spacing along the fin upper and lower surface is 0. . some of which are located on the interior of the fin. .001D. Like Part I. 2OD 40 20 30 1 2 3 4 5 6 7 8 9 10 M.90 so A70 .. some cells are arranged with the curvilinear grid lines not normal to the missile surface. Figure 69 Boundary for incipient separation at the leading edge of a blunt fin. the cells are distributed azimuthally so that the grid conforms to the upper and lower surfaces of the fins. the grid used in Part II is constructed such that the grid lines conform to the leading edges of the fins. Spacing at the body surface is the same as for the fine grid computations in Part I (0.00085D. Additionally. . 115 .. Unlike Part I. the cells are skewed substantially. Those points are essentially ignored during the computation. The author assumes that the current finitevolume methodology of handling gridcell geometric terms will minimize the error associated with skewing. and the flow solver is restrained from obtaining information across or through a fin. The node spacing at the finleadingedge/missilebody junction is 0. Because the fins now have thickness.000025D). To achieve a grid which is aligned with the fin leading edges. as proposed by [91]. the boundary conditions for the fin surfaces are enforced on the interior of the domain.
4. This extrapolation is given by: qi.& . All other boundary conditions remain unchanged and are described in Section 3.k = 2 qijm.xi. two point extrapolation is used.qijllz2. 116 . (58) where q is any conserved variable. the author believes that linear interpolation will provide an adequate inflow solution.2 Modification of Boundary Conditions. Linear interpolation is then employed to determine boundary data between points available from the solutions of Part I.k. The initial farfield boundary condition is also an interpolation of the solutions from Part I. Because the shape of the inflow boundary is chosen to correspond to grid points from the grid used in Part I.6. The solutions from the computations in Part I are used for the inflow boundary.6. From then after.maz.
The shock wave created by the fin leading edge results in a rapid pressure rise at approximately x/D = 20.7. First. and the accuracy of the thinfin assumption (Part I) assessed. Due to increased memory requirements. the effect of fin thickness on HART missile aerodynamics is addressed.78. Next. The nondimensional surface pressure along the centerline of the fins is used in this section to show the effects of the shocks and expansions that take place near each fin. and the fin trailing edge in a plane that intersects each fin's centerline. Secondarily. The surface pressure is measured along the missile body. the structure near the missile body and along the fin leading edges is shown with contour and velocity vector plots. The pressure is then relatively constanit along the fin's leading edge for the windside fin (4 = 135"). Again this data are compared to the results from the thinfin investigation. Finally. 7.1 Flow Structure Changes Near Fin The flow expands and compresses several times around each fin. The effects of fin thickness are assessed by examining changes in the flow structure near the fins and the drag and pitchingmoment coefficients of the missile. All cases are for clippeddelta fins and turbulent conditions. The impact of any differences found is evaluated by reexamining the drag and pitchingmoment contributions from the fins. = 2 and Mo = 6) and two angles of attack (a = 0* and a = 5*). the fin leading edge. the structure near the fin/body junction is explored through the use of streamlines and surface pressure plots. and rises gradually for the leeside fin 117 . Figure 70 shows the surface pressure plotted against the axial coordinate for M.. Results and Conclusions from Thickfin Investigation In Part II. thickfin effects are determined for only two Mach numbers (Mo. = 6 and a = 50. the inviscid structure is examined by measuring the surface pressure on the missile body and both fins. the drag contribution from the fins is computed and compared to the semiempirical relations used in Part I. The strength of the shock/boundarylayer interaction for the thickfin geometry is gauged by comparing this data in the fin/body junction region with similar thinfin data. The primary intent of this part of the research is to determine whether the pitching moment of the HART missile is significantly influenced by the thickness and crosssectional shape of the fins.
= 450).95 and two sweep angles (300 and 700). the shock/boundarylayer interaction influences the pressure upstream of the fin/body junction. Figure 72 is a plot of the surface pressure along the curve of intersection joining the fin and missile body for the thickfin calculation at Moo = 2 and a = 50. A sudden expansion at the clipped portion of the fin is evidenced by the fall in pressure at x/D = 22. the fin in [41] is cylindrically blunted and the current fin is flatfaced. The streamlines depict an attached boundary layer. and is attributed to the more abrupt expansion around the leading edge of the flatfaced fin. = 0. The author attributes the small differences in magnitude to these geometric differences.xi. Results are presented for both the windside and leeside fins. The pressure is nondimensionalized by the pressure 5 fin thicknesses upstream of the fin leading edge. = 2. the 118 .+ w sin 4. Figure 71 shows the streamlines in the vicinity of the windsidefin (0 = 135°)/body junction for Moo = 2 and a = Y. the flow recompresses downstream of the fin. as was the case for all thinfin computations. The boundary layer near the fin/body junction is not significantly affected by fin thickness. however. fin crosssectional shape. it is observed that the boundary layer remains attached for the thickfin geometry. Specifically. Finally.). which are extracted from the threedimensional solution in the plane that intersects the centerline of the windside fin. The streamlines are constructed from the axial and radial velocity components (u and v cos 4.(. and body curvature. a flat plate in [41] replaces the curved missile body used in the present computations. P. Exact agreement between the present results and those from [41] are not expected because of differences in Mach number. An attached boundary layer is also found at the leeside fin (4 = 450. The data from [41] are for Mo. The flow continues to expand along the top of the fin. not shown). The upstream influence through the boundary layer leads to a gradual pressure rise ahead of the fin/body junction (x = 0) =x for the thickfin geometry. Also. then rapidly expands again at the base of the fin (x/D = 23. A sharp drop in pressure is seen in the current data around x .18). This bluntfin/flatplate data is included in Figure 73 with the present thickfin results. Through examination of the velocity field near the fin/body junction. Clearly. Although the boundary layer does not separate. The upstream influence of the thick fin is also evident in data obtained experimentally for a bluntfin/flatplate geometry [41].38.
I .0 21B 22. . . . . . J 1 . I . .5 23. . . . Surface pressure coplanar with fin centerlines (Moo = 6 and a =5*) 1194 119 ..5 21.5 24.0 22.15 pf P 0 .0 23. .0 Figure 70. 1 20..
Streamlines in the plane of the windside fin (M = 2 and a = 50) 120 .Jr Jr Jr Jr Jr Jr Jr Jr 4 4 Jr 4* 44 Jr Jr Jr 4 Jr * 4 Jr Jr 4 4 4 4 4 44 4 44 4 Jr 4444 4 444 4 44 Jr Jr 4 'Ji ii .9 a 4 4 I '' Ii II' Jr 4 4 4 4 j4 I J I a' I I I / Jr Jr 4 4 4 4 4 4 1 II I / a' I I / II 4 / 4 4 1 1 Figure 71.' .
However. At MM.upstream extent of the pressure disturbance for the thickfin calculations agrees well with the similarly swept fin from [41].013* 0. and an empirical correction for these edges was not added to the thinfin results. The pressuredrag calculations from the thickfin tests are in reasonable agreement with the semiempirical data used in Chapter 4.007. 7. = 6.042* 2.632 0.0 6. since the azimuthal distribution near the fin is similar for thinfin and thickfin computations. the agreement is much closer when only the contribution from the upper and lower surfaces is examined. but now the coefficient for the thick fins is 0.011 Cd (total) 0. respectively.265 + (thick fins) 0. This is not surprising. f C+ + Cd (thin fins) 0. The turbulent viscous drag differs by a slightly larger amount because the leading edge and trailing edge contributions are included in the thickfin calculations. Table 17. the wave and base drag due to the fins is nearly identical at M. The thinfin result is 0. Drag components from thickfin computations (a = 0) A (thin fins) 0.042 and the thickfin result is 0.041. = 2.262 121 .026 Cd(thick fins) 0.0 0. the viscous drag coefficient is 0.619* 0.019 0. In fact. The pressure drag (wave drag plus base drag) and viscous drag is tabulated to allow direct comparison to the semiempirical data used in Part I.045* * See Table 5.041 0.006.011 for the thin fins and thick fins. Also included in Table 17 is the total drag'obtained by adding the thinfin results upstream of the thickfin domain to the thickfin results.006* C) otal 0. Again the coefficient for the thin fins is 0. The agreement between the computed pressure drag on the thick fins and the semiempirical drag on the thin fins suggests that the semiempirical relations give adequate estimates for the drag.2 Analysis of Fin Drag The fin drag from the thickfin computations (a = 0* only) is summarized in Table 17.006 and 0.
45"):. 1 2 3 4 Figure 72. a5* Cumn thick fin(4135): A70".50 225 P2.30. 41): A . a 5 2CQia tbick fn(4.80 1258 1. Pressure near computation fin/body junction: experiment versus thick. a5" 2.135): Aw0".fin 122 .1. 41): A .75 1. aS* 2.00 BhMfin/ihaPIa (Rd.75.25 1.75 BIMfW/flatpha (def. aS" Cu dukfinr(%.00 3 2 1 0 XXI.80 1.45): A70". 3 6 Figure 73.00 1.75 1 Curz fin(41rn 45:): A a706.00 3 2 1 0 1 xx. a5" 1.135):. Pressure near fin/body junction: thinfin versus thickfin computation (M = 2 and a = 50) 3. a5" • Cnunudda fin %. a5" Cuwnu ddckfin(4. Aa70". A0".
respectively. for the thickfin and thinfin computations. Complex flow structures are clearly evident below each fin. the bleeding around the thick fins is caused by the turbulent transport of momentum in a region of blocked flow. supports the categorization of thickness as a secondary effect on pitching moment.. = 6. the extent of the structure below the windside fin is greater for the thick fin than the thin fin.7. from the thin fins. with increasing Mach number. For the leeside fin.o = 2 and a = 5*. Figure 74 is a contour plot showing the pressure in the cross plane (yz plane) for M. Two relatively small effects of fin thickness are discernible in Figure 76. although a secondary effect. the structure below the leeside fin is more fully developed in the thickfin case.. The crossplane plot is for z/D = 21. thickness has a growing contribution to C.. Figures 76(a) and (b) show the crossplane velocity components at Mo = 2 and a = 50.. The two data points from the thickfin tests are obtained using the thinfin data upstream of the thickfin domain. The impact of fin thickness on Co is also relatively small. The level of correlation between Co from the thick fins and C. Although the influence extends further away from the fins in the thickfin case. bleeding of the pressure took place around the leading edges of the thick fins. for the sharpnosetip missile. The velocity vectors are also included in Figure 76. and the thinfin results from Chapter 4 are connected by a cubic spline curve. The value of C.9% at M. Figure 58 is duplicated in Figure 75 for convenience. changed 4. For the windside fin. = 2 and 31.9% at M. the flow is blocked by the fin surface and the expanding flow near the body. The experimental data are represented as open symbols. Also.3 Impact of Fin Interaction on Static Stability As with the turbulent. Like the thinfin configuration. Figure 77 is a plot showing the variation of C.8. First. the flow is blocked by the fin surface and the symmetry plane. The similarity in the thickfin crossplane streamlines with those for the thinfin case suggests that the structure of the flow is dominated by the blockage phenomenon.. Therefore. 123 . thinfin tests in Chapter 4. This is the same location used for the contour plot of pressure near the thinfin (Figure 58). and is a point near the middle of the fins. Comparing Figures 74 and 75 shows the bleeding of highpressure to the leeward side of each fin in both cases. the differences in the pressure along the fin surfaces are relatively small.o with M.
75 (Mo = 2 and a = 5 0) 0.00o0.25 1.50 0.1.50 2.500 0.25 0.00 0.75 1.250.00 0.25 1.250 0.5 z Figure 74.50 22.75 0.50 0. Pressure contours for thickfin model 1.00 0.00 1.25 50) Figure 75.50I 1.75 z 1.00 0. Case S28 (M.75 1.25 y 0.25 0. Pressure contours for thinfin model.o = 2 and a 124 .00 0 0.
50 0.00 02525025 ?fl H 0.00 0. 1.25  ~ > 0.' 0.75 z 1.25 Figure 76.75 z 50)..00 y 0.. 125 .00 1.25'' y0.00 0.25 0.25 0.00 1.25 0.50 0.75 0.1. = 2 and a = (b) thin fins (Case S28) (a) thick fins.25 I) 0.25 () 1..75'.75 ~ V 0./ .75 0. Crossplane velocity components (M.
smoo= body xpaimmt: sbap nmetip. = 6. In contrast.. the shock/boundarylayer interaction at the fin/body junction is very weak. is larger.. at M. thick fin  50  40Cu 30 20 10 1 2 3 4 5 6 7 Figure 77. body CCurm: bzrbukat thin fin Currnt turbulem. these two competing factors nearly offset each other. The clearest indication of the weakness is that separation does not occur at either Mach number tested. the flatfaced fin design allows an additional contribution to the pitching moment from the fin leading edges in the thickfin tests.. At M. clippeddeltafin model: thickfin results versus thinfin results Several factors contribute to the differences in the thinfin and thickfin pitching moment.4 Summary and Conclusions of ThickFin Analysis Due to the highly swept configuration of the fins. In ad126 .. and the change in Cm. is small. the thick fin occupies some of the body surface that was used to calculate Cm. the pitching moment created by the upper and lower surfaces of the fins is found not to depend on fin thickness. in the thinfin tests.70 A so a o * Exmimt: sharp noip. Cm. for sharpnosetip. Pitchingmoment coefficient. The contribution from these surfaces differed less than 2% between the thinfin and thickfin computations. 7.= 2. Interestingly. the leading edges provide a much larger component and the impact of fin thickness on Cm. Second. First..
between the thickfin and thinfin configurations supports the categorization of thickness as a secondary effect on pitching moment. Changing the leading edge shape through rounding or sharpening will undoubtedly reduce this drag component considerably. Although the fins are highly swept. This could be accomplished in much the same way that the pressure drag from the leading edges )isdetermined. 55]. 99. The strength and extent of the flow structures that develop in the blocked regions appear to be enhanced by fin thickness. It may be possible to empirically model the pitching contribution from the fin leading edges. = 2 and 10% at AfM= 6. 127 .. As with the thinfin tests. the pressure drag is still significant because the fin edges are square. this blockage phenomenon causes bleeding of pressure around the fin leading edges. The contribution to the total drag is 6% at M. The change in pitchingmoment coefficient for the thickfin tests is caused by lift on the leading edges.3. As stated in Section 7.dition. The structure of the flow near the fins is affected by the turbulent transport of momentum in regions of blocked flow. The level of disturbance found in this investigation is similar to that found by other investigators on simpler geometries [64. the disturbance in pressure is limited to one fin thickness upstream of the junction.. An empirical approach to modeling thickness effects is attractive because of the dramatic reduction in memory requirements for the calculation of the entire missile flowfield... the level of correlation in C. The author speculates that this is caused by further constraint of the flow geometrically dictated by the physical thickness of the fins.
total drag is compared to experimental tests of this shape (with fins) [93]. an initial assessment of the algorithm is made assuming axisymmetric flow.Appendix A.78 and compared to experiments [47]. At the same time. tangent ogivecylinder model with a flat base. timeintegration technique. tests are run for Mach numbers ranging from 1. Calculations are made on axisymmetric bodies with fineness ratios between 9 and 23. Roetype. Since the ongoing goal of the research is to predict the aerodynamics of both finned and unfinned bodies. The cylindrical form of the NavierStokes equations is described in Section A. The fineness ratio is 23. and Mach number is varied from 2 to 7. turbulent flow is computed on pointed. LID = 14. finstabilized missiles. Algorithm Validation Using Axisrymmetric Equations The intent of the validation experiments is to determine the algorithm's accuracy and sensitivity. several different nosetip shapes (Figure 78). at Mach 2. flat. with varying degrees of bluntness. Laminar flow is also computed on a hemispherecylinder. and then establish the grid requirements for HART missile research. Therefore. and spherically tipped tangent ogive shells.5 to 8. and Reynolds numbers from 1 x 106 /m to 2 x 10/m.95.1. laminar flow is computed on a highly blunted. incorporating an upwind. at Mach 7. Specifically. Then. threedimensional equations (Chapter 3). These comparisons provide a refined level of flow interrogation to verify the algorithm's accuracy. this method must be both frugal in its computer utilization and accurate in predicting the aerodynamics of long projectiles. Both laminar and turbulent conditions are simulated on the forebody and base. LID = 9. The current research investigates the robustness and accuracy of the basic algorithm for the flight conditions of interest. fluxdifference splitting (FDS) scheme [71]. the ability of the algorithm to calculate a complete configuration is paramount. These requirements must be met so that the approach can be effectively used to study the stability of long. slender. To establish a baseline. To demonstrate this ability. one objective of the current study is to build a computational methodolgy that is sufficiently general to compute accurately a wide variety of flow structures. 128 . The NavierStokes equations are solved with an explicit. are investigated. In Addison. This provides an appropriate reference point for future investigations using the full. These results are compared to both experiment [13] and independent computational results [29].
Experimental/computational models and nosetip shapes 129 .Highly Blunted Tangent OgiveCylinder HemisphereCylinder Pointed Tangent OgiveCylinder Spherically Blunted Tangent OgiveCylinder Flat Tip Tangent OgiveCylinder Figure 78.
Ar and IW. A. transformation would yield aU + at + ap  + H= aR + S  + W. The first procedure reformulates the governing equations. transferring the curvilinear terms to sourcelike vectors. Accuracy and sensitivity issues are summarized in Tables 18 and 19. Numerical accuracy is judged from comparisons with theoretical solutions for stagnation pressure and velocities in a turbulent boundary layer. Both oi these methods are explored here. Recent work reported in [63] proposed two methods of handling this singularity for threedimensional grids.1 Axisymmetric Equations The geometric singularity associated with axisymmetry poses significant difficulty when employing either twodimensional or threedimensional grids.5 and 7. Some differences are found on both the forebody and base. pv 2 . described later.These tests determine the algorithm's ability to predict accurately the influences of upstream effects on boundarylayer development. in the context of axisymmetric flow. the effect of the entropy correction parameter on surface pressure is assessed for both the nose and wake regions.F). Finally. which incorporate the axisymmetric assumption. Details are highlighted in the sections that follow. J (60) (61) (62) "J [pv.1 Reformulation of Finitedifference Equations. 38. 31. 1 (riE+ 7.0. The baseflow tests are performed to demonstrate the validity and accuracy of the wakeregion turbulence model. These results are compared to experimental data reported in [10.1.F). (e + p)v] (63) 130 . 53. A. For the cylindrical form of the NavierStokes equations. turbulent flow is computed for a flat base region and Mach numbers between 1. These are highlighted in subsequent discussions. 11. Additionally. 663. puv. (59) where 0 E F 1 = = = = j1u. ji'(E + .
As/D Wall Spacing Ay/D Wall Spacing Yalwall Entropy Func. cl MinMax Values 61121 61181 0. Perf.85 Difference in Result 0. cl Entropy Func.624. 53. 11. Gas Theory i Configuration HC PTOC & BTOC PTOC & BTOC CFB Case Description Case 5 (see Table 20) Y+. Summary of comparison between computation and theory or experiment Basis of Comparison Tip Stagnation Pressure Streamwise Velocities Surface Pressure Integrated Base Pressure Comparison With Inviscid. 31.1% 15% 28% 4% Tested Configuration HC HC HC HC PTOC & BTOC CFB CFB Notes Laminar Laminar Laminar Laminar Turbulent Turbulent Turbulent I Table 19.85 Quantity Observed Cd Pressure One Body Diam.04 0. 38. Coefficient.0 0. Line. Coefficient.0020.001 0.Table 18. Summary of sensitivity evaluations Parameter Varied Number of Nodes Along Body Number of Nodes Normal to Body Spacing at Sing. From Surface Tip Stagnation Pressure Tip Stagnation Pressure cuhhlwall Base Stagnation Pressure Integrated Base Pressure Change in Quantity 4% 0.00010.85 0.0050. 1 (see Table 19) 121 Nodes Along Body and y+ : 1 31 Nodes Along Base and cl = .5% 0. 66] 131 .1% 2% 0.0020.2% 4% (max) 1% (avg) 10% (max) 1% (avg) I Notes Laminar Turbulent Turbulent Turbulent I Law of Wall/ Law of Wake [101] Experiment [12] Experiment [10.
The grid points themselves represent the vertices of the cells. (64) Since Hf and W are not evaluated on the singular (symmetry) line. a pure. For the second procedure. 132 . (rJ')E+ i 1 (65) 1 &)F. This approach is considered a hybrid finitedifference/finitevolume approach by the author. threedimensional. which was finitedifference. P. Unlike the approach taken in [63]. 0.)F. As before E and F are the inviscid Cartesian flux vectors. This is important at computational boundaries. finitevolume representation of the geometric terms is utilized. Expressions for the geometric terms appearing in (65)(68) are calculated at cell interfaces from grid cell vertices (grid points). local physical conservation is addressed when applying (59) to a cell in physical space by calculating the flux terms and geometric terms at the faces of the cell.2 Governing Equationsfor Finitevolume Method.)E + (rJ'7. Although the transformed equation deviates from its integral formulation roots. and W are the generalized viscous flux vectors and the viscous source term. S. . [0. H= ]T. The geometric terms (metrics and Jacobian) associated with (60)(63) are calculated at cell centers from finitedifference expressions. but now S= = (rj')U. but H differs. and the reader is referred to [91] for an excellent treatment of the subject (expressions for threedimensional grids are given in Appendix B). respectively. The other vectors R.1. F and the viscous flux vectors can be found in [1]. Again (59) applies. they are never singular. where the boundary conditions are handled in a manner consistent with a cell interface representing a boundary. The explicit forms of E.and E and F are the convective flux vectors in Cartesian coordinates. and averaged to obtain values at cell interfaces. ji1 is computed from the following expression: j' = x~r.zrc. (rJ (66) (67) (68) P = (rJ'i?. These evaluations are straightforward for twodimensional grids. A. the current technique implements the discretization of (59) with the finitevolume methodology.
33]. only the terms involving metrics differ.j + Ej) + rj'&. The flux is chosen so that the scheme is secondorder accurate in regions of smoothness and firstorder at points of extrema (e. (F+. this is discussed in Section 3. For both procedures. Discretization of (59) is as follows: A.).ri _~' • . The exact form of E and F are provided below. denoted by 01+ + = are ajai+½j)(gi'+Ij +gj a133 !(i½ • l+j + (71) 133 .3 . shocks).. The convective fluxes are typically determined at cell interfaces in two ways. R. The elements of the vector ti+L . The explicit.).) (69) ] _ _..+•j = 2 1(. This scheme achieves secondorder accuracy by applying Roe's firstorder.+ 4 (70) A similar expression is obtained for the hybrid finitedifference/finitevolume approach. the viscous terms are centrally differenced.Discretization of the NavierStokes Equations. using either MUSCL or nonMUSCL extrapolation.+.+• (E.g. secondorder formulation employed for this study is based on the work of Yee and Harten [104. while the convective terms are upwind differenced. The current scheme uses the nonMUSCL technique.j + Fi ) + (rJ')._. The formulation is a modification of Roe's firstorder algorithm.+i.. When using finitevolume expressions for the geometric terms. and has been referred to in the literature as a modifiedflux scheme.j) interface is approximated by E. The vector 0 acts to limit the characteristic variables.+$jt. fluxdifferencesplitting algorithm to a modified inviscid flux.+j ].+.1. thereby providing higher accuracy.'.1.t i At 2i+ (F i+½j . the convective flux in the ccoordinate direction at the (i + I.
twopoint extrapolation is used. as follows: g•+=j = minmod(al +Lja+ (74) or +. a. . where T is an entropy correction to IzI.sgn(o[+½)a!+a amin [Icj+ 1])]. freestream conditions are applied. (72) if at i+40 = 0..1 fC419 0. The characteristic variables. For the body surfaces.)[max ( i 1. and either constant wall temperature or an adiabatic wall is used. noslip is applied. and secondorder expressions for the finitedifference grid. zero normal pressure gradient is used. the current study uses the minmod function. 0. the boundary conditions are very straightforward. 134 .4 (here with Roe's average [71]). (73) Finally. Finally. Due to the supersonic condition at both the inflow and outflow. (75) A.4 Boundary Conditions.(= sgn(mi+n. centerline symmetry is enforced using ghost cells for the finitevolume grid.1.)('1.+ji and U. At the inflow. gi. is given by o(z) = ZA 2). and at the outflow. The function.Uj).where a+ is the characteristic speed evaluated at some average of U. are given by = R+j(Ui+lj. It is the same as (31) in Chapter 3. The characteristic speed is modified by y which is defined by = 2 ooa+.+½. aj+½i. 9. for the limiter function.
onedimensional..A. without additional loss of generality: Ui+2 Ui2. it is assumed that a local extrema did not exist.0 Istorder +Ivl 135 .(76) 5i ax (77) Implementing the current algorithm gives: u+'= U" AX+ ( IAIJt (Un. AA? (ui+i  Ui"1)AtBu! (78) 2• AX AX AA secondorder term 2 u U!. a scheme developed for a homogeneous system may not be stable.i + firstorer term The above form of the equation applies away from discontinuities. it is assumed that the algorithm does ui _> ui. Therefore. Using a Fourier stability analysis [1] produces the following relationships for the amplification factor. inhomogeneous equation of the form a+ of Linearizing the equation yields Ott + A Ou= Bu. Since the purpose for this stability analysis is to investigate limitations due to the source term. Furthermore. these assumptions should not hinder the analysis. 20 + un. G: ?nd order G = 1. .5 Stability Analysis for Source Term.1. Consider a scalar. Near a discontinuity the secondorder term is zero. no entropy correction function). a simplified stability analysis was done to anticipate any stability restrictions..1 _> ui+1 _> _Ž not modify the wave speed (e. Therefore.v(isin#1) +Iv( \ G=Iv(isin)  )+lvl(cosa 1) + lvI(1 &( " B + I(cos'3 . nonlinear. Since a source term appears in the governing equations.g.1)  ( lv1) cos 2. For the limiter evaluations.
the value of A+ is 26 for a smooth. 1 produces the following: Enforcing IGI _< ?nd order 1st . The nondimensional 136 .22. The well known and widely used zeroequation BaldwinLomax turbulence model [5] is incorporated for the forebody boundarylayer simulations. is given by the Prandtlvan Driest formula Pi = p(icyD. v = i and IGI 5 1 for stability.4). as B gets very large. as r + 0. A. the eddy viscosity. both reduce to Imd < 1. B may get large relative to A for this study. the stability may be adversely impacted. This shows that the source term should have negligible impact as the grid is refined.)2 1•w. (79) where w is the vorticity.. is the van Driest damping factor. y is the normal distance from the surface. The boundary layer is divided into inner and outer regions. (80) The superscript "+" in (80) denotes a nondimensional variable. X is the von Karman constant (0. Referring to (59)(63).jlul JAI ) Ivl < 2 + JAI + 2Ax< 0 i1] IvA I I+ JA < Note that for B = 0. Clearly. 1.6 Turbulence Modeling.where. pi. In the inner region. flatplate in incompressible flow [101]. which is given by D = 1ex p ]. due to the 1/r dependence of f. nonporous. and D.order lJl3 . which is the familiar CourantFriedrichsLevy (CFL) condition [1]. The actual effect of the source term in the secondorder formulation is plotted in Figure 79.
Cd0 •. the eddy viscosity is defined by p.5(Cdebi/l. 28. The second alteration improves the two coefficients in the outer region. denoted by A+.. is given by F. The Clebanoff intermitancy factor..distance._ is maximized over the I values at each station.. where F. 137 . The Clauser constant. C• and These modifications are also outlined in Section 3..... The nearwall region. flat plate [101].. Ckleb. Both N and Z are defined in Section 3. 1011. is redefined as A+ = 26N.. = [1 + 5..4 by (39). The turbulence model switches from the inner to the outer formulation at the first (closest to surface) value of I for which pi > •o. = pC. (84) Again. y+. = max(l1wID•. N is the pressure gradient contribution. Two modifications prove important. and I..). F. is 2PtWIWkI Y =/Y (81) The subscript "w" denotes conditions at the wall or surface. and Z is the compressibility contribution. Cjb. is the value of I corresponding to Fma.0168.4.)6 ]1 (83) The constant.. nonporous. the nondimensional distance 26 is the incompressible value for a smooth.Cl. In the outer region of the boundary layer. and the value for the Cp is modified from the incompressible value (explained below).FF. The first changes the van Driest damping factor to account for compressibility and pressure gradients [49..Z. is also modified from the incompressible value and is explained below. is 0. (82) The function F.....
7 Computer Resources All grids are obtained using GRIDGEN [80]. and the nose radius is varied through the Reynolds number. This technique was originally applied to subsonic flow in [541. it is difficult to apply a conventional zeroequation turbulence model. = lm. A. A firstorder TVD scheme does not predict the stagnation quantities as well. Although highly empirical. 61 x 61. Surface clustering is employed in the generation of all grids. The freestream velocity is varied through the Mach number. nor does 138 . Both finitevolume grid types (cell interface along singular line). The method is adapted. nearly stagnant region.1. Figures 81a and 81b are the pressure and density contours for Moo = 5 and R. Turbulent viscosity is determined from (42) in Section 3.1 Adiabatic Wall Results. for supersonic flow in this investigation. and its validity tested. So. The method has the advantage of speed and low memory requirement. and representative of the secondorder accuracy. a strictly empirical relation is used to compute turbulent viscosity in the wake. A. The accurate calculation of stagnation quantities behind the shock. Freestream conditions representative of flight at an altitude of 30km are used.For a flat base with a large. and the precise prediction of the expansion around the sphere are illustrative of the accuracy.1 Spherical Nosetip. A. for the hemispherecylinder is shown in Figure 80. an elliptical grid generator developed for Wright Laboratory.2.2 Integration of the NavierStokes Equations for a Laminar Forebody Numerical simulations are performed for several geometries and flow conditions to assess the general applicability of the methodology developed. OH. and finitedifference grid types (cell center along singular line) are employed. A. The solution fidelity is quite good. the validity of this approach is shown through numerical examples. WrightPatterson Air Force Base.1.2.2. A typical grid.
5 10.0 7.300 Figure 79.0 12.0 Figure 80.050 0.800 .5 15.0 0.0 2. Effect of source term on stability 12. 61 x 61 grid for hemispherecylinder 139 .850 0.000 .5 5.5 x/D 10.100 0.250 0.0  7.0 2.0.000 0.200 0. 0.1.5 5. 0.950 0.5 0.
the Reynolds number based on nose radius is ReR. = 24. The stagnation point heating is within 15 percent of the experimental value of Rose [72]. Also the pressure around the body agrees well with the experimental data of Oliver [62]. Freestream conditions representative of flight at an altitude of 30km are maintained. Again the solution fidelity is very good.2. and the growth of the boundary layer around the body. the same level of agreement with the experimental data of Oliver is achieved using the coldwall boundary condition. Since surface pressure is perhaps the most important computed quantity. Since no chemistry effects are included in the numerical calculations.it predict the expansion as precisely [57]. This is an excellent test for the adiabaticwall condition. this level of agreement is good.66cm are selected to make direct comparisons with experimental data [72].1.2 Cold Wall Results. Figure 84 shows the heat transfer distribution 140 . the adiabaticwall temperature is the stagnation temperature (TIW = To). For Pr = 1. = 10. Away from the shoulder.5 percent of the value predicted from the normalshock relations. this is examined in detail. Figure 82 is a plot of the nondimensional surface pressure around the body. A surface temperature of 300°K and a nose radius of . A. When Pr was set to 1.400. the computed temperature was constant to within 1K and very close to the stagnation temperature predicted by the normalshock relations (within 100). The stagnation pressure is within 0. but provides some further measure of overall accuracy. Note the smooth expansion in the inviscid region. Heat transfer is much less important for the current study than surface pressure or skin friction. the heat transfer distribution can be well modeled with a cost 0 profile [2]. Figure 83 is a density contour plot showing the large density gradients across the boundary layer. At these conditions and M. No appreciable change in the surface pressure from the adiabatic tests is noted. Of primary concern is the surface pressure.
10 1. .17 4318 z 6D 1.17 O m08 1. altitude = 30km.Im.1. Nondimensional pressure and density contours (M. adiabatic wall) (a) pressure. (b) density .p7 000 i 0. 1• 141 .5.M R Figure 81."7 LU M80 1.00 4.3 (b) I.
01 .25 0.AS .00 0 25 50 6 75 I 100 Figure 82.75 0.  .30kmi.001 a A c ommm Obw 0.420 470 40 OmO Figure 83. altitude ./R = 0..0. Density contours: MO = 5.1.8 ( = is the body inclination angle given by Figure 16) .50 Pstag  0.66cmn. and cold wall 142 .70 .U r 0.= 5.1.0 where 0 0. Surface pressure versus angular position: M.
: Cf. Cf. the skin friction should exhibit a linear proportionality with distance around the body.T 2f o. Both of these trends can be seen in Figure 85./Pr) (87) These relations are only slightly modified when computing Cf..) are usually based on local information at the boundarylayer edge. the order of magnitude agrees with other computational results on similar geometries [2].an*(88) 1.~ =.~ rPu[(1 =w CH. 143 . U2' u petie(hawh) or and CH. However. P =)Re.3 Skin Friction and Heat Transfer Coefficients..) .2.'" e) I (1) M' f].= q (85) Ch ~ ' G.1. Note the excellent agreement with the cost 0 rule. Perhaps equally important is the trend in skin frictior.Pr) CH• (We [(r) 1 + 'II MM2) ..2o (86) + YM.T.nondimensionalized by the stagnation point heating. (2. On a loglog scale.(1 . A. approaching zero at the stagnation point [2]. i•* (1 /)(•Y Pr 1M' _u 2 (89)T (89) A reliable set of experimental data for skin friction could not be obtained by the author. The heat transfer coefficient should be nearly constant over a fairly large portion of the nose region. and CH. . Local skin fric tion and heat transfer coefficients (C 1 . The following relations are used to compute Cif and CH. and CH.
ReD = 259. ? = 144 . Computed skin friction and heat transfer coefficients: M. Heat transfer distribution.= 7. R 5cmn.106 10*' 0 25 so e 75 100 Figure 84. computed versus theoretical: M. and cold wall = 7. altitude = 30km. 5cm.. altitude = 30km. and cold wall 100 " 100 Figure 85.000.
laminar flow is assumed in all simulations for this configuration. hypersonic. and 0. Measured surface pressure and impact pressure are accurate to within 2%. The results are shown in Figure 86. the freestream test conditions correspond to Re = 9. but unexpectedly poor results at some other conditions. that develop at higher Reynolds numbers. Table 20 summarizes the 145 . The experimental model was 19. = 700*K. The variation in stagnation heating clearly follows the ' theory. The experimental data for flow over a hemispherecylinder [47]. A.4%.1..01%. shock slope and position are accurate to within 3% [47]. the reservoir pressure was 2.2.2 HemisphereCylinder (HC). respectively.0001D. the supply temperature was 645"K ± 0. and impact pressure.A.17 x 10 6 N/m 2 ± 0. 0. (59)(63). For the experiment. In other words.78.1 mm in diameter (Figure 78). At Mach 7. Again. Note that a finer grid is required to model accurately the thinner boundary layers. and the Mach number was 7.2. The behavior of "acomputational model when various parameters (e. wind tunnel located at the California Institute of Technology. Calculations are done with three grids of successive refinement. and 121 x 121. all solutions are obtained using T. The experiment was conducted in the continuous flow. Since a well known trend in stagnation point heating exists (qw ca 7X). The three grids are 61 x 61.4 Effect of Nose Radius on Stagnation Heating. local time stepping. with spacing near the surface equal to 0.2 x 10'/m. closed return. surface pressure. Com puted solutions are compared to experimental estimates of shock location. the computational procedure may give remarkably good results at one set of conditions.g.01D. provides a more complete basis for evaluating the accuracy of computed solutions both near the body and away from the body.7 ± 1%. 91 x 91. Except where noted. the nose radius is systematically varied to test accuracy in this somewhat broader sense. and the cylindrical transformation.001D. M and Re) are changed is "a good indication of its accuracy.
103
"
1011
o S 1213121 61XG1
100 1
.. .
.
. . .1
.
.
,
.
. ... .. 1.
. . ...
10"1#106
167
Re,
Figure 86.
Effect of nose radius on computed stagnation heating: M.,,= 7, P 5cyn, altitude =30krn, and cold wall
146
case history of numerical experiments. The distance along the body is s, and 0 is the angular spacing near the symmetry line. The grid type [finitevolume (FV) or finitedifference (FD)J has an undesirably large influence on the solution character near the stagnation point. Again, a FV grid has cell interfaces on the symmetry line, while a FD grid has cell centers on the symmetry line. As seen in Figure 87, the FV grid produces a local aberration in the pressure, manifested as an overprediction in stagnation pressure. It is important to note that this error is not the same as distortions in the stagnation region reported in the literature as "carbuncles" [63, 43]. This phenomenon does not appear to begin at the shock, and is not accompanied by recirculating flow as with the carbuncle. The entropy function appears to prevent effectively the carbuncle anomaly. Since similar calculations, using MUSCL extrapolation to achieve higher accuracy [43], do not demonstrate this error, the author believes nonMUSCL extrapolation accentuates the geometric singularity when using a FV grid. According to Palmer and Ventkatapathy [63], the "key to effectively treating the singularity line boundary lies not in boundary conditions themselves, but rather in the proper determination of the metric and flux terms on the singular line." This observation is underscored by the results of this work. A different form of extrapolation (nonMUSCL versus MUSCL) is used in the current scheme, as well as cylindrical versus Cartesian equations. None the less, either technique works well when using a FD grid (Figure 88). For the remainder of tests on this geometry, and the rest of the validation study, a FD grid is employed. Since the entropy correction function is not used in the boundary layer on the nose and forebody computations, its magnitude primarily affects the solution away from the body. As previously stated, the entropy correction function is made large enough to prevent nonphysical solutions; increases beyond that point affect stagnation pressure less than 0.2%. The gridrefinement experiments (Cases 16) are evaluated by examining the stagnation pressure as a function of the node spacing at the surface, A (y/D)min,
147
1.00
0.75
•
FhiaDiffamc= Grid • .Fmiv•Vohm Grid
0.5
025
0.00 1 0.00
0.25
0.50
0.75
1.00
I/D
Figure 87. Aberration in surface pressure near the stagnation point (a = 00)
1.00
0.75 FinizVohum Formuilaio   Cylindical Tnmfondmia

0.50
0O25
0.00. 0.00
025
0.50
0.75
1.00
xID
Figure 88. Comparison of surface pressure for modified solution procedures
148
and the node spacing at the singular line, A~min. The computed stagnation pressure is compared to the value predicted from inviscid, perfect gas theory. The computed stagnation pressure accuracy depends primarily on the grid spacing, along the body, near the symmetry line; the error decreases as this spacing decreases. The effect due to spacing at the singular line is shown in Figure 89. The computed pressure (nondimensional) appears to approach asymptotically the theoretical value of 0.925. The values differ by about 2% for the coarsest grid and less than 0.5% for , 0.570 
(see Table 19). In contrast, stagnation pressure error is virtually unaffected by grid refinement normal to the body, and varied < .1% for the range of spacing tested, .001 (see Table 18). Although not compared to an exact so.0001 < A (y/D)min _< lution, surface pressure away from the stagnation point exhibits identical sensitivity to grid refinement. A comparison with experiment is shown in Figure 90 for Cases 1 and 7. Agreement within experimental accuracy is achieved. The computed pressure contours illustrate that shock location and shape are also accurately predicted (Figure 91). A notable exception is the slight over prediction in standoff distance, which the author attributes to the perfect gas assumption. Finally, impact pressure profiles examined at a station downstream of the nose, i/D = 3 are given in Figure 92 for Cases 1, 3, 7, and 10. There are two important results from the comparisons of the impact pressure profiles. First as with surface pressure, grid refinement does not appreciably alter the solutio, the body. Second, although undocumented
by the experiment, an adiabaticwall condition matches the experimental data best. Again, agreement within the experimental accuracy is achieved. A.2.3 Highly Blunted Tangent OgiveCylinder (HBTOC). This is the longest
model used (LID = 23), and is particularly blunt in comparison to other tangent ogive models tested. Its length, highly blunted nose, and flat base make it a demanding choice for fullconfiguration analysis. To increase the demand on the algorithm, the solution is impulsively started from freestream initial conditions. These conditions are set to match those of the ballistic gun range testing facility at Eglin AFB. 149
GTS.0001.0100 0. GTS. FD Grid Global Time Step (GTS).0100 IAfmin 4.ni 0.80 0.75 . FD Grid 1. .. 1.I .0 Figure 89. .15 1.(dWg=e) 4.0050 0.0100 0.0 3.0150 0.0 2.0075 0.15 1.0 Sensitivity of stagnation pressure to refinement of node spacing near the axisyminetry line 150 .86 0.Table 20.00 0.0 A 0 .72 1. FD Grid Adiabatic Wall.95 Cue 6 0.0100 1.0100 0.57 1.0 // 0. FiniteDifference (FD) Grid 8 9 10 0.0100 0. 0.15 1.15 AyD= .15 0.15 CommentsI Ay/D = .001. FD Grid FV Metrics.0390 0. Summary of numerical experiments for hemnispherecylinder ICase] 1 2 3 4 5 6 7 Grid Size 61 x 61 91 x 61 121 x 121 121 x 121 121 x 121 121 x 181 121 x 181 121 x 181 121 x 181 121 x 181 A(s/D).49 1.0 5.
Surface pressure. Pressure contours.0 7. Case 1 and Case 7 Blowup of Nows iqiw 10.0 25 5.5 0.5 x/DY 10. 15.100 0 3 6 sID 9 12 is Figure 90.0 L re 1 '0 13rI'A 2.0a xanm Shock Sudfae U CAI 5.0 0.0 I Figure 91.0 12. Case 7 151 .
75 1 lz SAIs Oo ozs y.75 1 .0 125 Figure 92. (b) Case 3. Impact pressure (:ilD 3): 10 (a) Case 1.00 ON YrD)2 (un~bmm 0~o . (C) (d) C. 01..50 0. (d) Case 152 . rD2 (imba) o.(a) [(b) Egi Curimm dr "/ Lr' 10.0 * 0. (c) Case 7.
the experimental terminology is adopted. lower Mach number fluid near the stagnation point of blunt noses. Second. tip 'P' refers to the pointed model. under differing upstream conditions. strictly laminar flow is simulated on this configuration. First. the transition to turbulent flow is affected by the higher entropy.. The experimental models were 4.3 Computations for a Pointed and Blunted Tangent OgiveCylinder (PTOC and BTOC) in Turbulent Flow The turbulent boundary layer on a projectile or missile is affected by the nose geometry for several reasons. The reservoir pressure was 6.95 these correspond to Re = 6.3 x 107 /m. for the case of a flat nosetip. To distinguish between the nose shapes. Experiments are used for comparison [12]. 0.95 cm in diameter and are shown in Figure 78. high Reynolds number. a series of nose shapes are investigated to evaluate the accuracy of the current algorithm to calculate a developing boundary layer. During the Princeton experiments. The effects of the highly blunted nose are evidenced by the relatively long recovery of surface pressure to the freestream value.1. and tip 'R3' refers to the spherically blunted model.5 for 153 . tip 'F3' refers to the flat nose model.99 x 107/m is presented in terms of pressure contours and surface pressure in Figures 93(a) and 93(b) respectively. The experiments were carried out in the Princeton University 20 x 20 cm. blowdown tunnel [12]. A.A description of the system is given in [93]. Third. the flow separates. Therefore. the vorticity associated with shock curvature interacts with the boundary layer downstream of the nosetip.8 x 105 N/M 2 ± 1% and the supply temperature was 265"K ± 5%. Boundarylayer transition occurred at s/D .9 . To provide an unambiguous baseline for subsequent turbulent calculations. the wall temperature remained within 3% of the adiabatic value. supersonic. so standard sealevel conditions are chosen. At Mach 2. This is essentially an open atmosphere system. The computed solution for Mo = 3 and Re = 6.
154 .0. (b) surface pressure 00.10 0 (a) 10 101 10'100 xID Figure 93. tangentogive/cylinder (M". a~ 5612 grid points): (a) pressure contours. = 3. 5 30 3 Solution for highlyblunted.
However.0. Downstream of this point. It is well known that Roetype approximate Riemann solvers. and R3 respectively. can calculate accurately boundary layers with relatively coarse meshes [90].1 for tip F3. since the BaldwinLomax turbulence model was developed to produce law of the wall/law of the wake profiles. Velocity error throughout the boundary layer is ±0. where s is the distance along the surface from the stagnation point. and at sID . The pressure rise due to the recompression shock 155 . and sID = 1. The performance of the algorithm is not adversely affected by the abrupt change in effective viscosity (a small variance is detectable in the pressure at the surface). 1. Rather. the turbulence model is used. This is an appropriate comparison.7 . the gridrefinement analysis concentrated on the appropriate spacing nearest the body. the timeaveraged velocities in the turbulent boundary layer are compared to the law of the wall/law of the wake. and F3. P. The accuracy in the surface pressure measurements was ±1. Therefore.tip P.0. For 1+w . sID = 1. Figure 95 shows the comparison for the first six pressure taps on the nose of tips R3 and F3. Also shown is the computations of [30] for tip R3.0%. a point of transition to a fully turbulent boundary layer is specified. For turbulent flow computations. which include information about all waves.1. Surface pressures are computed to within experimental accuracy. Boundarylayer transition is not simulated.1.. the velocity gradients at the surface vary significantly with grid spacing normal to the surface.7 for tips F3. there are no further variations for grid refinement equivalent to 1wall <1. a comparison is shown in Figure 94 for tips P. However. see Table 18 (1+is the distance from the body surface nondimensionalized using shear velocity as given in [5]). R3. at sID .2.8.5 . The current calculation shows expansion on tip F3 slightly sooner than the experimental data. These "transition" points are s/D = 0.5 for tip R3.2% (see Table 19). variations in the data due to changing stagnation conditions over the duration of the experiments showed a stagnation pressure variance of about ±3%.
The boundary layer thickens quickly downstream of the transition point. The vorticity contours clearly depict that the shock layer and boundary layer merge. Results from [30] agree qualitatively with the current results using the original BaldwinLomax model. Since boundarylayer development is sensitive to a pressure gradient. The small depression in the velocity profile is further evidence of the interaction. The author attributes differences between the current calculations using the original model and those of [30] to differences in transition modeling. F3. however. The extent of the separation region is approximately I caliber. Figure 96 shows computed velocity profiles for tip R3 using the original BaldwinLomax turbulence model and the modified model. Computed profiles from the modified model are also shown for tips P and F3 in Figures 98 and 99. The computations of [30] are also included for comparison.is accurately predicted. Although separation was not clearly visible in the experimental data. is visible in the streamlines produced from the current calculations (Figure 100). this result agrees both qualitatively and quantitatively with the calculations of [30]. Excellent agreement with experiment is again obtained. the modifications to the turbulence model described earlier prove important in this region. the original BaldwinLomax model and a transition model (polynomial function) were used. The arrow in Figure 97b shows the approximate point above the surface where this merger can be seen for x/D = 3.5. respectively.ts 1. transition was initiated at s/D z . This is discernible from the vorticity contours (Figure 97a). and fully turbulent conditions were achieved at s/D . 156 . Significant improvement with the modified BaldwinLomax model is visible. The vorticity associated with the bowshock interacts with the rapidly thickening boundary layer near the ogivecylinder shoulder. Separation on the flat nose tip. In [30].5. with the exception of the final station on tip P.26. The author has no explanation for this isolated discrepancy.
2.0
1.0
Is
0.5
,S U
2.0."
1.0
:
€
2.0
zo
4
tip P, (b) tip R3, (c) tip F3
157
15.0
(a)
12.5
10.0 . 7.5 ..... u
s.0
2.s
0 .0
0.25 0.30 0.35 0.40
. . .
. ..
..
0.45
0.50
x/D
15.0
12.5
10.0
,
7.5
a.S
•CW"
0.0
. .
. .
. . . . . . .
. . . . ..
i .
..  . 
.    
025
0.30
0235 x/D
0.40
0.45
Om.5
Figure 95.
Surface pressure in nose region of a tangentogive/cylinder (M, = 2.95, a = 00): (a) tip R3, (b) tip F3
158
0.10
0.08
o.o6
" oCurrent (Original D.L Model)
Current (Modmfied B/L Model)
0.02 0.00
0.10 (b
0.05
o.o

Current (Modified BIL Model) Current (Originl BL Model)
0.02 0.00
0.10c
0.02
0.02.10
Figure 96.
Velocityv profiles for tangentogive/ cylinder (M,,,, 2.95, R3): (a) x/D = 3.26, (b) x1D =5.05, (c) x/D = 6.33
ct
00, tip
159
2.0
(a)
1.5
".1.0
0.5
0.0 0.25
0.75
1.25
x/D
1.75
2.25
2.75
3.25
0.14 0.12
(b)
O.l0

en (Modified B/L Model)
0.06
0.04
0.04
.00 0.91
U
0.92 0.93 0.94 0.95 0.96 0.97 0.96
0.99
1.00
1.01
U U.
Figure 97.
Vorticity interaction on a tangentogive/cylinder (M. = 2.95, a = 00, tip R3): (a) vorticity contours, (b) velocity profile at x/D = 3.26
160
102 0.00 0.26.. tip P): 161 .02 0.33 2. Velocity profiles for tangentogive/cylinder (M.00   Curfewt (Modified B/L Model) Comaud *Experiment 0.06 0.02 0.00 0. a 00.05. (b) x1D = 5. (c) x1D = 6.10 0.6  Current(Modified B/L Model) 0.0 Figure 98.06 0.10 0. (a) z/D = 3.95.901..04 0.10 (b) 0.0.
10 .O Figure 99.02 0.. Model) 0. tip 162 .00 0.04 0.6 (a) 0.. Velocity profiles for tangentogive/cylinder (M.0.06 0.33 a = 0°.95.00 0.10 0. (c) z/D = 6.06  Current (Modified BA.02 0. (C) 0.05. F3): (a) z/D = 3.Cuirt (Modified BI.02 0.26.10 0m. (b) z/D = 5.06 (b) 0. = 2.
425 0.12 0.475 0.450 xID 0.14 0.400 0.16 0.13 0.0.13 ~0.14 0. Separation on flat nose tip (tip F3): (a) velocity vectors.15(b 0.16 0.500 Figure 100. (b) streamlines 163 .15 0.
is taken from theory [101]. a more rigorous approach is not attempted. ReL = 2 to 9 x 10i. The boundarylayer solution. 11] are all for pointed. Over the range of Mach numbers. The integrated pressure coefficient compares very well with the experimental data (Figure 102). the wake turbulence model accurately 164 .A4 Calculationsfor a CylinderFlat Base (CFB) Configuration Using an Empirical Turbulence Model To establish the validity of the wakeregion turbulence model. Due to the empiricism of the turbulence model. Upstream of the corner. and turbulent flow at the body/base corner. The computational model has a small cylindrical section (. the pressure still varies across the base an order of magnitude above experimental measurements. The computed solution at the rearward stagnation point exhibits behavior similar to the forward stagnation point. (42) is implemented. The exception is [38] which had fins and 6.. The effect of changing the entropy function on the total or integrated base pressure coefficient is small (about 4%). the Reynolds number is held fixed at ReD = 5 x 105. In spite of its empiricism.85.05D) prior to the flat base allowing upstream influence of the corner expansion within the boundary layer. the modified BaldwinLomax model is used. local aberration in pressure results in an overprediction at the stagnation point. and they remain very geometry dependent. 66. Although some improvement results from using a FD grid type. a universal function does not yet exist. all base flow numerical experiments were conducted with f = 0. 53. the analysis is done with the base flow isolated from the forebody.5 < LID < 30.85. 38. Downstream of the corner. specified at the inflow boundary of the computational domain. unfinned bodies with LID =z 10. cylindrical. For all bodies. Therefore. 10. much better behavior is seen (Figure 101). It is very important to keep in mind that entropy corrections manifest themselves as additional artificial dissipation in the algorithm. The compiled experimental measurements [31. In addition. By increasing the coefficient in the entropy correction function to 0.
Two techniques dramatically accelerate convergence to steady state. The correction norm is defined by IIAUII= )  2 (u. Pm•n = 0.01 poo and Pin = 0.simulates the macroscopic nature of the flow. on a CRAY YMP8. This additional restriction on the base is continuously enforced. during the transient flow development. Typically. is about 9 minutes on a CRAY YMP8.000 points). per node point. The convergence criterion is the reduction of the correction norm. 165 .)] (90) where n represents an arbitrary time level.01 p 0 0are enforced. First. approximately 110 words of memory are required per node point. 5 microseconds of CPU time are required per time step. Also. the Reynolds number is reset to the correct value. by 5 orders of magnitude from the initial conditions. For a turbulent solution. Total computing time. and the subscript I denotes one of the four conserved variables. Second. the Reynolds number is artificially set to 1 x 1010 at t = 0. A. using the convergence criterion related above.5 Convergence Acceleration Techniques for SteadyState Computations Local time stepping is used to accelerate the solution to steadystate. After the shock wave moves away from the body (outside the boundary layer). a lower limit on density and pressure is used to prevent negative values on the base. or 2. The convergence history shows the robustness of the present method (Figure 103). but actually only necessary for the first few hundred iterations from an impulsive start. IIA UII 2.4 megawords for the finest grids used (22.
. Re=30to35 x 10 Ref.Re= 5 to 9 x 10' Ref. Re=2x 10' Current..50 p/ p. O. 5. .00 . Computed versus experimental base pressure coefficient 166 .10 0.30 e a 0. .0. Influence of entropy correction parameter on computed base pressure (M.40  0.= 2. 10.31. . 53 & 66.00 Figure 101.  "Ref. Re4 t8x 106 Ref.20 0. Re4tox 106 Ref. a=0) 0.50 0.0 M.A .00 0. 0.1 1. 0.25 . .0 Figure 102.0 3.002 0. . 11. 38 (w/fins). .Limiting Pressure (Vacuum) 0.0 7.U 0.U.0.75 1. Re=5 x 10' C 02.3 .
10' 100 10*1 10 2 10 10 4 4 S1o 10. . I.12 0 . 10*6 . 10"10 1 10" 10. .I 2000 4000 6000 8000 10000 Locaw Time Steps Figure 103. Convergence history for forebody/base combination with local time stepping 167 . .
both in terms of computational rate and memory requirement. The algorithm's robustness is enhanced by artificially increasing the Reynolds number and controlling the boundary conditions on the base during the initially transient flow development. viscous flows. cornerpoint expansions. When using nonMUSCL extrapolation to obtain fluxes. and flexible numerical scheme for the simulation of highspeed. and turbulent boundary layers. accurate. The effectiveness of this approach is demonstrated by the algorithm's efficiency throughout the range of conditions tested. Modifications are primarily made to the treatment of singular lines. The results bolster the primary purpose of this investigation: the de P ment and validation of an efficient. 168 . baseflow separations and associated recirculations. While completing the validation. These improvements are reflected in the analyses of threedimensional flows in Part I and Part II. The axisymmetric algorithm is the foundation for the construction of the threedimensional algorithm described in Chapter 3. including the case of a complicated boundary layer/Lhock interaction associated with the separation near the corner of the flat nosetip. The general applicability of the method is validated through the accurate calculation of a wide variety of axisymmetric flow structures: detached bow shocks.A. some notable complexities arose that required the development of important modifications of previously published techniques. the wake turbulence model and the entropy correction function of Yee. An aberration is encountered near stagnation points that is different from the previously documented carbuncle phenomenon. The complicated turbulent boundarylayer development is accurately calculated for various nosetip shapes with a modified BaldwinLomax turbulence model. Results are in close agreement with published experimental estimates (within experimental accuracy). recompression shocks (both in the wake and at a separated region near the nosetip).6 Conclusions from Axisrymmetric Validation The current algorithm represents an efficient implementation of finitevolume methodology.
THIS PAGE Is MISSING IN ORIGINAL DOCUMENT . /69 .
In particular. a nonconformal grid structure is applied to a combination axisymmetric compression ramp and axisymmetric expansion ramp. These computations serve two purposes: assessing the ability of the present method to capture the azimuthal variation in the flow. The added equation necessitates a rederivation of the generalized inviscid flux Jacobians and the associated eigenvectors. the threedimensional algorithm is applied to an ogive/cylinder forebody at a small angle of attack. First. These computations are performed to determine the accuracy of the method developed to compute the flow near a surface (Section 3. the differences in the algorithm which arise out of the extension to a higher dimension are outlined. Finally.3). The extension of the algorithm for threedimensional flow requires that further validation be performed. This represents a small simplification to the algorithm. Second. Next.1 Modifications to Algorithm for Threedimensional Equations The governing equations for a threedimensional flow are described in Chapter 2. a comparison with previous results is made for the ogive/cylinder body analyzed in Appendix A. there is no source term associated with the threedimensional form of the equations.1. and determining the sensitivity to azimuthal refinement. Validation of Threedimensional Equations The axisymmetric form of the algorithm is capable of determining the many aspects of the general behavior of the current method. There are three differences between the axisymmetric form and the threedimensional form of the governing equations that require the algorithm to be modified. First. 170 . Then.Appendix B. the effects of dissipation and the influence of grid refinement along the body and normal to the body are addressed in Appendix A for several forebody configurations in both laminar and turbulent flow. Finally. This is addressed below in Section B. B.1. an additional momentum equation is added to the set of conservation equations.
The generalized inviscid flux Jacobian for the cdirection is now given by [25] 0  G U~ts Ucv (1 . and 4z are computed from the projected cellface surface areas from: =Sy and C.UcH #34v + 4w WvU.1. B.1.1. Also. with V.the geometric terms that correspond to projections of cell faces into the Cartesian directions become more complicated.Uw iUo . They are discussed in Section B. . or Wc. H is the total enthalpy and Uc is the contravariant velocity given by U. (91) ff. where yis the ratio of specific heats. and "y j= (y l)(u + v +w). The terms G.fi)&4U + Uc N3u +v /4u + Gw G f34v + Gu (1 1 3 4W + GU /O3w •±v 0 /34 )•v% + UC + H4 = /3 63.w+ uc IwUo + H4 #uUc + H4...2. 171 . and replacing U.. (1 /3)4.2).1 Generalized Jacobian and eigenvector. = 4u + + 4w.1. (92) where 1ý is some average cell volume at the interface that St is evaluated (Sf is defined in Section B. QU . 4. . 4v The generalized inviscid flux Jacobians for the other two directions have similar construction and are obtained by replacing ý with q or C.
. respectively.kP kU .1. (96) 172 . + k'P k.k'C Vkc Wk~c sk. p.. and k.)p sk. The terms k.u k.v)p H + UOc H .k~u)p 9k. =S! s V) (94) (95) The normalized contravariant velocity. + (k.k'P kV k.. the terms are defined by [91] S 4378 = (F4 X) (r8 ). The subscripts R and L denote a value to the right and left of an interface. is averaged using the following [25]: p = V/•p7. U6. + (kv .(...u + kYv + kW B. + k. k. + kP ku + kVp kV. are defined by 4: = $S" (93) ky = k:.k.. k..2 Geometric Terms of a Threedimensional Fluid Cell...The associated right eigenvectors are [25] k. 1. +(ku where S q2 1 .'C .. .= k. + kC W+k'c u ..ci The density. The geometric terms for a finitevolume methodology are obtained by using the two diagonals for each face on a cell. For the typical cell (Figure 104).. is given by C.k•P kW  .
r3) x V4 . . The pressure and boundarylayer velocity profiles are directly compared to assess any differences. §C= V( . the axisymmetric results from the supersonic. A twodimensional grid is constructed 173 . The computation is performed in only half the azimuthal plane by enforcing a symmetry condition (Section 3.2 Application to Azisymmetric Baseline Validation of the threedimensional algorithm is begun using the results from Appendix A.r2).4). B. Typical sixsided fluid cell from a structured grid S123 415 = = S = (r. turbulent flow over a spherically blunted tangentogive/cylinder are used for comparison. Figure 104.1 5 2 . Specifically.r5) (97) (98) x(vs  rl).
This is the same procedure used to obtain grids for the analysis of the HART missile in Part I and Part II.95). The velocity profiles show that the growth of the boundary layer is also the same as the axisymmetric computations (Figures 107a107c). This grid is rotated about the body centerline to produce a threedimensional grid with dimensions 61 x 201 x 33 for the half plane.3 174 . Nonconformal Grid Applied to Azisymmetric Compression Ramp The grid used in the analysis of the HART missile does not conform to the fin leading edges (Section 3. twodimensional grids are built (61 x 81) and rotated about the body centerline to produce the threedimensional grids. The axisymmetric solution is converged such that the residual error is several orders of magnitude less than the convergence criteria outlined in Appendix A (Section A.with spacing normal to the body identical to that used for the axisymmetric calculation. The pressure comparison shows that the threedimensional algorithm predicts essentially the same surface pressure as the axisymmetric algorithm (Figures 105 and 106).5). The agreement in pressure and boundarylayer development indicate that axisymmetric flow is accurately calculated with the threedimensional algorithm.000025D). (0. Both a conformal grid and a nonconformal grid are built and used with the threedimensional algorithm. = 2. To validate the method developed to handle the nonconformal grid structure.3). Also. Only very small differences in the pressure are discernible. turbulent flow is computed for a combination axisymmetric compression ramp and axisymmetric expansion ramp downstream of an axisymmetric ogive/cylinder forebody (M. Again. The level of "nonaxisymmetry" in the threedimensional computations is seen to decrease as the solution converges. the computation is performed in only half the azimuthal plane by enforcing a symmetry B. The small difference in the boundary layer from the threedimensional calculations is due primarily to different levels of convergence. The two grids are shown in Figures 108 and 109.
0 I I I 0.0 7.3D eqoaim (*. Axiymmetm .5  3D equdum 07() 3D ueatim (. I . 025 0.0 2.0 .0) 3D]equ) m ( W(=8o) 1.30 0.5 Figure 106.0 12.0 0.0.50 Figure 105. 1.5 0.15..2.0 2.40 0. = 2.95 and a = 00 2.5 5.0 P 7 .45 0..5 p.5 x 5. PP. .95 and a = 0* 175 .35 x 0. Surface pressure near the nosetip of a tangentogive/cylinder in turbulent flow: M. I I .5 0. Surface pressure away from the nosetip of a tangentogive/cylinder in turbulent flow: M..IW) 10..5 2..
05 3D equations (=0) e= 180) 3D equations ( (a) 0.000.AXiSYmmTD?2riC (b) 0.05 qaon 3D equations (# =) 3D quations(* 180) .10  3D equations 00) y_.8~5 ..00 0.0.70 ' U 0. Velocity profiles for a tangentogive/cylinder in turbulent flow with M. (c) .= 6.05 3D: equations (t=180") C) 0. 1.26. .00 0. 05 1.05..33 176 ..1o y D 0.15 0. (b) • = 5.95 and a = 0°: (a) L =3. = 2.0U1.15 0.00 Figure 107.00 10 0. D 0.1O y D 0.0U N0 U.15 0.00 0.
Results using this grid will be referred to as "finegrid" results. Results from these grids will be referred to as "coarsegrid" results. the accuracy to which the surface pressure and the turbulent boundary layer are calculated. Figure 110 shows that the extent of the disturbance upstream of the compression ramp is slightly underpredicted by the nonconformal method. corresponding to angular spacings of 11. The pressure in the separated region and at the expansion ramp are equally well predicted by each method. For the half plane. The grid in this part of the validation is the same as that described in Section B. Again the surface pressure and boundary layer are compared to values determined experimentally [13].2. The nonconformal grid is intentionally coarsened to test the limitations of the methodology. is determined. For the fine grid.81250. for a forebody at some nonzero angle of attack.250 and 2.4). Given the coarseness of the nonconformal grid. the computed pressure on the leeside (4 = 00) and windside 177 . The number of nodes in the azimuthal direction is varied from 17 to 65.90. both the conformal and nonconformal grids have dimensions 61 x 81 x 33. The refinement study is made to determine the sensitivity of the current method to azimuthal spacing. the level of agreement between the conformal results and the nonconformal results is believed to be acceptable by the author.4 Calculations for an OgiveCylinder Forebody at Angle of Attack (Turbulent Flow) Next.condition (Section 3. The tangentogive/cylinder body used in Appendix A and Section B. Separation occurs prior to the compression ramp for all three computations.2 of this appendix is pitched 2. B. Grids with slightly coarser spacing near the surface are also implemented to conduct an azimuthal refinement study. This angle is selected to compare with experiment. The conformal grid is also used with the axisymmetric form of the algorithm to provide additional data for comparison. respectively.
Nonconformal grid for axisymmetric compression ramp 178 . Conformal grid for axisymmetric compression ramp Figure 109.Figure 108.
00 0. 3DaWm.0 // 01 / 2  Axiq. the velocity profiles on the leeside and windside are compared to the experimentally determined profiles (Figures 113 and 114).. However. oaftcmal aMca4an• "r 225 2. The point which separates the laminar region from the turbulent region is not varied around the body because this information is not available from the experiments. Agreement within about 10% is seen.1800) of the ogive/cylinder body is compared to the experimental pressure in Figures 111 and 112. . boundarylayer transition is not uniform around a body at angle of attack [79]. Also. is the same as the point designated in Appendix A and Section B. for the fine grid.95 (4.75 20.0 22. For the computations. the point used to separate laminar and turbulent regions.7).25 / 1.0 Figure 110.0 23. and the author believes the discrepancy be 179 ..2 (= 1.75 2. Surface pressure from computations for an axisymmetric compression ramp: Moo = 2. caofrml Vrd srid•dd 3Dqap.0 24.0 21.2.00  1. Agreement within the reported experimental accuracy is achieved.
In addition. B. Only the coarsest angular spacing gave different results. the separation point and reattachment point for turbulent. Calculations at angle of attack show fairly good agreement with experimental data. and downstream of the attachment point. The solutions from from the grid refinement study also showed little variation in surface pressure (Figure 116).tween the computations and the experiment are a result of neglecting the azimuthal variation of transition point.. The velocity profiles for the solutions from the grid refinement study showed that prediction of the attached boundary layer does not depend on the angular spacing. The impact on pressure is small. The accuracy of the method proposed to handle a nonconformal grid structure is also demonstrated. the pressure within the separated region. Results for a = 00 agree with previously validated axisymmetric results (shown in Appendix A). respectively. The results from the grid with coarse and refined spacing near the surface are compared in Figure 115. azimuthal grid refinement is shown not to affect the solution significantly. Computed surface pressure is within the experimental accuracy and computed boundarylayer velocity profiles are within 10% of experimental. The leeside and windside profiles near the ogive/cylinderbody junction are compared to the experimental data in Figure 117 and Figure 118.95 with a = 0* and a = 2. The azimuthal variation of surface pressure is well predicted by the present method. Mach 3 flow over a cylinderflare configuration compares well with computations made using a conformal grid.9*.5 Conclusionsfrom Threedimensional Validation The implementation of the threedimensional equations is demonstrated for flow on a tangentogive/cylinder forebody at M. 180 . Also. agrees with the conformal computations. = 2. over the range of spacing tested. In general.
0 2.0 P7...40 0. i . 0 0.5 A.A  £ ..9' 181 . .5 • * . .95 and a = 2.. . .5•• 2. .5 Figure 112.0 025 .   . . 0. : *=i80 1.A  . Surface pressure near the nosetip of a tangentogive/cylinder in turbulent flow: M. I .5 0.5 p. I 0. 1. .0 Curet: 1w( Crren. .0 I 2.8 12.. . .18.5 I x 5..50 Figure 111.0 I 0. .45 0.90 2..95 and a = 2. . n I . .5 *Epiet  current$0# Canme .30 0.35 x 0.. I . = 2.Currment =w 1ao" 4I 10.5 5.o xpaiment *=o 2.0  "..0 7. .0 12. . Surface pressure away from the nosetip of a tangentogive/cylinder in turbulent flow: Mo = 2.
= 2.u' ' u 0. 1.= 5.90 1. . . Figure 113. (c) 182 . 0o20 (b)  0.15 r 0.05.05 020 (C) * 0.o 0.020 (a) * 0. .90: (a) .10 0.05 Es o .90 .05 U.05 0 .95 and & = 2. *.15  Eperint Current r o.uo.9o.75 U 0.6. o.00.75 .33 = 3.05 °. 0.10  Experiment cufrent 0.1o 0.7 U 0.15 *Experiment Current r 0.9o 1. Leeside velocity profiles for a tangentogive/cylinder in turbulent flow with Mo.05 V.• . (b) ._0 0.26.
u 0. 0.9 1. (b) .26.10 (C) *Experiment Currmn r D0. 0.05.05 Figure 114.75 o 0.33 183 . (c) S=6.10 * Experuncnt Cuffent r 005 (b) .05 U.= 5.95 and a = 2. Windside velocity profiles for a tangentogive/cylinder in turbulent flow with Mo.10 *Exparfimt Current r (a) 0.o0 0. = 2.0.05 .90: (a) M = 3.75 U 0.0.0o 0.05 °.90 1.75 V.90 1.05 U.6.
75 . .95 and a .... d# ..4181 ..25.eside Wimdle  2. I I . 0 30 60 90 120 180 160 Figure 115... 1.. Comparison of the azimuthal variation of surface pressure for azimuthal grid refinement: Mo.. 2. ... ....2..625" C.75 __P  "E.. d#.d# 11.538 Exp...1.Computed.95 and a = 2...481 1.846 .90 184 . d*.625 1..2..'D0.00  .625" Culued....25 * Exp. 0.. ..... .5...7. xD6. "id...1..2... Y/D .. ..75.Lsd 'Wimkide  2.50...00 * 0.50 p . Exp..25 .....00 SExp.513 Exp..80 1. 2. .00 1. ..xD" • •'"""Exp.513 p P1... xiD3.. 0 30 60 90 120 150 180 Figure 116. . L... .  xAD 0. ".p..25' ....90 2.812!5 1..846 ... uD6..'...... = 2. YD.. .. Exp.COaaqUd.5.538 . ... K... d# 5..50 . ...25 1.. xzD a 3... 2.. Comparison of the azimuthal variation of surface pressure using coarse and refined spacing near the body: M.
60 0. Leeside (4 = 00) velocity profiles for azimuthal grid refinement: Mo = 2.75 I .61 x 81 x 17 grid 61 x8lx33 grid 61 x81 x65 VWi 0. and & = 3.020 0.15 * Experimn= .26 185 .00 0.10 0.95.90.05 Figure 117. a = 2. u 0.05 0.90 1.
00 I . a = 2. = 2.60 0.05 0.9°.75 u 0. Windside (4' = 1800) velocity profiles for azimuthal grid refinement: M..26 186 .95.0. I I 0. and M = 3.10 Rxperinnt 61 x 81 x 17grid 61 x 81 x 33grid 61x81x65 VW  0.90 1.05 Figure 118.
Appendix C. Table 21. Resource utilization. FL) "* Cray C90 (Vicksburg. MS) The usage of these resources is normalized by Cray YMP8 equivalent central processor unit (CPU) hours and summarized in Table 21. Computer Codes and Resources The following hardware resources were utilized to perform the computational portions of this dissertation "* AFIT Stardent ST2000 "* AFIT Convex C220 "• Cray YMP8 (Eglin AFB. Computations Code Development & Axisymmetric Validation (Appendix A) 3D Validation & Thinfin Investigation (Appendix B & Part I) Thinfin Investigation (Part I) Thickfin Investigation (Part II) Machine AFIT Stardent ST2000 Utilization (CPU hours) (normalized to Cray YMP8) 61 AFIT Convex C220 226 Cray YMP8 Cray C90 473 40 187 .
0 2. along with a description of the subroutines contained therein. each code encorporates the BaldwinLomax turbulence model. Optional missile fins are modeled with an infinitely thinfin approximation (i. The grid construction and boundary conditions are described in Chapter 3. using boundary conditions applied on the interior of the domain). and was used for the validation tests in Appendix A.. AFITENSAXI Version 2. 188 . AFITENS3D Version 1.1 All three codes allow Euler or full NavierStokes calculations using upwind differencing of the convective terms and central differencing of the viscous terms. A short description of input/output files for the above codes and a short synopsis of the subroutines employed is provided below. AFITENS3D Version 1.C. This version was used for the validation tests in Appendix B and the thinfin HART missile simulations in Part I.0 uses the finitevolume method on a single grid block with an HH topology for a complete missile configuration.1 allows the missile fins to have thickness.1 was used for the thickfin HART missile simulations in Part II. AFITENS3D Version 1. A short synopsis of the computer codes employed is provided. Also. again utilizing an HH grid topology. Three main codes were employed: 1.0 3.e. Finally. AFITENSAXI Version 2. AFITENS3D Version 1.0 is an axisymmetic code employing a finitevolume methodology. AFITENS3D Version 1.1 Software Documentatijon for General Algorithm This document was prepared using Latex on an AFIT Sun SPARC 10.
direction gradient .dat restart.determines the time step using the linearized stability restriction etasweep . bcstag4 . and pitching moment "*Subroutines bcaxi .applies boundary conditions at bisymmetry plane for FV grid bcstag3.input file to specify CFL.builds the numerical fluxes and performs the operator in the v. Re. drag.dat . etc. M. and drag forces and the pitching moment efgvect .applies boundary conditions at bisymmetry plane for FD grid bcaxi3.restart file for conserved variables "*Output Files restart. farfield.dat . outflow.drv .applies boundary conditions on domain interior for missile fins for FV grid cd .computes the axial.constructs the inviscid and viscous flux vectors eigdt .AFITENSAXI/AFITENS3D "* Input Files afitens.dat .grid file grid.applies boundary conditions at inflow.constructs a threedimensional FD grid gridstag . outflow.applies boundary conditions on domain interior for missile fins for FD grid bcstag . normal. surface for FD grid bcaxi2 .restart file for conserved variables forces.constructs a threedimensional FV grid 189 .applies boundary conditions at inflow.. .computes velocity and temperature gradients gridaxi . bcaxi4 . surface for FV grid bcstag2 . a. farfield. lift.lift.
A comparison of computer code performance is shown in the following tables.calculates the geometric terms using a finitevolume method turbvisc . AFITENS3D Version 1. zetasweepB.0  /home/fluids/kmoran/3D/THIN 3.computes eddy viscosity xisweepA./home/fluids/kmoran/3D/THICK C. and the associated drivers.0 on several different machines. AFITENSAXI Version 2. xisweepB .0 .lamvisc . restart files. and grids.2 Algorithm Performance and Memory Requirements The current threedimensional solver (AFITENS3D Version 1. zetasweepD . Table 23 shows the data processing rate of AFITENSAXI Version 2./home/fluids/kmoran/AXI 2. Resonable computations rates are seen for several AFIT workstations when they are used in a parallel mode. 190 . are archived on the AFIT Sun SPARC 10 (enterprise) in the following directories 1.builds the numerical fluxes and performs the operator in the zl2norm . AFITENS3D Version 1.0) requires ap proximately 80 words of core memory per grid point.calculates the residual norm between time steps ( direction The computer codes.computes molecular viscosity metrics . The current technique is very efficient in terms of CPU usage. Table 22 shows a comparison between the axisymmetric version of the current solver (AFITENSAXI) versus axisymmetric solvers used by other researchers.1 . zetasweepC.builds the numerical fluxes and performs the operator in the Cdirection zetasweepA.
Type Implicit B/W Implicit B/W Explicit FDSRoe Implicit B/W Explicit FDSRoe*** Machine Type* _ _ DPR** 1.92 x 10i 2. Comparison of computer code performance for axisymmetric solvers.0 6.09 x x x x x 104 Comments Thin Layer Eqns 2D MUSCL 2D NonMUSCL Cray XMP Cray XMP Cray XMP Cray XMP Cray XMP/216 105 104 10105 Data processing rate (DPR) in seconds/iteration/node point ** AFITENSAXI Version 2.5 7.0 1.0 Table 23.96 x 10' 8.Table 22.50 1.93 1. Developer/ User BRL WL WL AFIT AFIT ** Algorithm.0.5 2.60 2.0 4.75 x 101.32 x 101 13. Factor over CrayYMP8 DPR* 81.41 x 105 5.0 5.15 1.09 x 101 1. Comparison of data Machine Type Sun SPARC 2 Stardent ST2000** Titan 3000 Convex C220*** Cray XMP Cray 2 Cray YMP8 * ** processing rates for AFITENSAXI Version 2.34 x 101 Data processing rate (DPR) in seconds/iteration/node point Run in parallel using four processors * Run in parallel using two processors 191 .3 4.
0 192 .40 x 105 1.90 X 104 5. Algorithm Type Implicit B/W Explicit FDSRoe Implicit FDSRoe I Cray 2 Cray YMP8 Cray YMP8 Developer/ User BRL WL WL Machine Type I DPR* 1. Table 24 gives a comparison of the threedimensional solver developed for this dissertation with those of several other researchers.00 x 104 I Comments I PNS MUSCL MUSCL NASA Ames AFIT * ** Implicit FDSRoe Explicit FDSRoe** Cray YMP8 Cray YMP8 2. Comparison of computer code performance for 3D solvers: full NavierStokes and turbulent flow.50 x 10i 1.Finally. Again.50 x 101 MUSCL. Table 24. laminar NonMUSCL Data processing rate (DPR) in seconds/iteration/node point AFITENS3D Version 1. excellent performance is seen with the current method.
and Richard H. "A Comparison of Finite Volume Flux Vector Splittings for the Euler Equations. Assistant Professor.. and H. An Introduction to Fluid Dynamics.25. G. Beran. December 1993. D." AFATLTR778. "Aerodynamic Characteristics of 2. Aerodynamics of Wings and Bodies." AIAA Paper 780257. An Investigation of the Aeroof a Series of ConeCylinder Configurations at a Mach dynmic Characteristics Numer of 6. Anderson. MS Thesis. 10. Lomax. NACA Research Memorandum L51J09. An Analysis of Base Pressure at Supersonic Velocities and Comparison with Experiment.. and Landahl. 3. Raymond A. Philip S. John D. 1967.. Graduate School of Engineering. 5. Dolling. Cooper.86. B. 2. Compilation of Wall Pressure and Turbulent Boundary Layer Data for Supersonic High Reynolds Number Flow Over a 193 . Personal Correspondence. 1965. Baldwin B." AIAA Paper 850122. G. 1951. C. Air Force Institute of Technology. Blake. Dean R. 8. S. 1951. Sears. and Robinson. L. 1989. 4. New York: Dover Publications. K. Anderson. Hypersonic and High Temperature Gas Dynamics.0. 7. D. Chapman. Ashley H. and J. NACA Report 1051.00.. Inc. K.. J. 1984. M. Inc. Series in Computational Methods in Mechanics and Thermal Sciences. S. Dale A. 12. Anderson. Tannehill. 0. 11. S. 6. John C. Batchelor. 3..6 to 4.75 at Mach Numbers from 0. New York: McGrawHill. New York: Cambridge University Press. S. New York: Hemisphere Publishing Corporation. Pletcher. and Pallas. 1992. "Thin Layer Approximation and Algebraic Model for Separated Turbulent Flow. Air Force Institute of Technology (AU). Computational Fluid Mechanics and Heat Transfer. 0. January 1985.and 4Caliber TangentOgive Cylinders with Nose Bluffness Ratios of 0. and 0. A Numerical Study of HighSpeed Missile Configurations Using a BlockStructured Parallel Algorithm. Thomas. Gray. Ralph D. 9. W. January 1977.. and K.6.. WrightPatterson AFB OH. AFIT/GAE/ENY/93D07. E. Butler.Bibliography 1..
Drew. Volume 17 Number 8: 525526. 22. August 1955.. "A NonParameterized Entropy Fix for Roe's Method. "Aeroballistic Research Facility Data Analysis System (ARFDAS).. P. Gnoffo." AIAA Paper 870280. I." Journal of Aerospatiale. "Pressure Drag Calculations on Axisymmetric Bodies of Arbitrary Moldline. Donaldson. 21. "Application of Program LAURA to Threedimensional AOTV Flowfields. Gaitonde. W." The Aeronautical Quarterly. B." AIAA Paper 910636. "Aerodynamic Test and Analysis of a Slender Generic Missile Configuration. 16. 17. NJ. et. May 1986. Etkin. January 1990." AIAA Paper 900280. Air Force Institute of Technology (AU). 1991. F. "Base Pressures at Supersonic Velocities. New York: John Wiley & Sons.. 19. and W. Dolling. R. "The Effects of String Supports on the Base Pressure of a BluntBased Body in a Supersonic Stream. 18. January 1986. A. Mechanical and Aerospace Engineering Dept. HighResolution TVD Schemes for the Analysis of I. 1972. David S. January 1987. M. A. "Experimental Study of Supersonic Turbulent Flow on a Blunted Axisymmetric Body. G. D. A." AIAA Journal. September 1982. 24. Report #1585MAE.Blunted Tangent Ogive Cylinder at Small Angles of Attack. 13. August 1950. Journal of the Aeronautical Sciences.. al. Gates." Reader's Forum. Fischer. H. School of Engineering. Driver. August 1989. Number 5: 793799.. Volume 24.. Cipolla. L. H.. S. Inc. and J. R. Princeton. Gnoffo.. and Yee. Dubois.. C. K. Eglin AFB. A. Gray." AIAA Paper 893368. "Enhancements to Program LAURA for Computation of ThreeDimensional Hypersonic Flow." AIAA Paper 860565. WrightPatterson AFB OH. P. 23.. Bernard. Gabeaud. December 1991. Capt Mark A. McCandless.. PhD dissertation. 221229. Hathaway. Winchenbach. Mehlman. "A Systematic Comparative Study of Several High Resolution Schemes for Complex Problems in High Speed Flows. September 1988. FL. and Jenn. S. A. 15. Princeton University. and G." AFATLTR8848. 194 . S. Dynamics of Atmospheric Flight. January 1991. R. Inviscid Supersonic and Transonic Flows IL Viscous Flows with Shockinduced Separation and Heat Transfer. 20. 14.
and Burnett. Volume 49: 357393. 34. "On a Class of High Resolution TotalVariationStable FiniteDifference Schemes.95." Presented at the 31st Aeroballistic Range Association Meeting. S. "An UpwindBiased. Granville. NACA Research Memorandum L52E06. P. 33. Granville. 27. "A Modified Van Driest Formula for the Mising Length of Turbulent Boundary Layers in Pressure Gradients. MayJune 1992. 1987. Dolling. "Aeroballistic Research Facility FreeFlight Data Analysis Using the Maximum Likelihood Method. A. PointImplicit Relaxation Algorithm for Viscous. Bernard J. FL. and P. Volume 111: 9497. Eglin AFB. Roger G. W. and Flat Tipped Shells at Mach 3." SIAM Journal of Numerical Analysis. D." AIAA Journal. S. Hathaway. 1959. Harten. R. 29.." AIAA Paper 900587. and David S. Hathaway. "High Resolution Schemes for Hyperbolic Conservation Laws. Effects of Stabilizing Fins and a Rear Support Sting on the Base Pressuresof a Body of Revolution in Free Flight at Mach Numbers from 0.. 26. South Burlington Vermont. Dolling. "HART 1993: Freeflight ARF Test Results.. P. Paul Weinacht. Paul Weinacht. "A Finite Difference Method for the Numerical Computation of Discontinuous Solutions of the Equations of Fluid Dynamics. 31. J. Hathaway. 35.3. Volume 25. A. 37. 1989. P.. Hart. Sb. Hyman. "NavierStokes Computations for Pointed.7 to 1. 1952. Volume 29. Lax.. 36. Spherical. Complessible PerfectGas Flow. A.. and Whyte. H. Number 12: 1624. Guidos." Communications on Pure and Applied Mathematics. "Comparison of NavierStokes Computations and Experiment for Pointed. K. W. 1984. Harten. Gnoffo. H. R.1627. A. California. Arrow Tech Associates. Harten. Volume 47: 357393. "BaldwinLomax Factors for Turbulent Boundar Layers in Pressure Gradients.25. December 1979. J. Godunov. Number 3: 305311. S. "On FiniteDifference Approximations and Entopy Conditions for Shocks." NASA TP2953. 195 . M. R." Journal of Fluids Engineering." Mat. 30. H. and David S. W. Bernard J. Whyte. 1976. Volume 21: 123. October 1980. Spherical and Flat Tipped Shells at Mach 2. 1983. 05403. "Projectile Design Analysis System (PRODAS). 32." Journal of Computational Physics." Test Report AROWH94001.. 28.. Volume 24: 297322.. H."Journal of Spacecraft and Rockets. Guidos. Pasadena. Fegruary 1990." AFATLTR7998.
. W. S." AIAA Journal. Journal of the Aero'autical Sciences. Kubota. R. M. Freeman K. 1982. H. Turbulence Modeling. "Simulation of Bluntfininduced Shockwave and Turbulent Boundary Layer Interaction. Toshi.. November 1985. 40.S. "Effects of Small Nose Bluntness on Static Stability and Magnus Characteristics of a Projectile Shape at Mach 0. Flight Dynamics Laboratory. March 1950. and Winchenbach. 46. CM. 49. 41. October 1978. "Base Pressures at Supersonic Velocities. and Stollery. Packard. Volume 116. Dash. Volume 154. Hung. Wright Research and Development Center. Gaitonde. G. "Description and Capabilities of the Aeroballistic Research Facility. Cranfield Insitute of Technology. Shang.91 and 3. J. Volume 16. D. January 1985. 48." AFATLTR8708. July 1986.. D. A Study of the Interaction Between a Glancing Shock Wave and a Turbulent Boundary Layer . May 1987. June 1957. P. and Miguel Visbal.03. James.. MD. C. Knight. Army Ballistic Research Laboratory/AMC." Graduate Aeronautical Laboratories California Institure of Technology (GALCIT) Hypersonics Wind Tunnel Memorandum No. April 1958." AIAA Journal. L.." Reader's Forum. Interim Report. 43. L. Hussain. Volume 17 Number 3: 185187. S. ppl53185. "Investigation of Flow Around Simple Bodies in Hypersonic Flow. Hill. C. 42. Eglin AFB. BRLMR3535. "Nonequilibrium Hypersonic Flow Solutions Using the Roe FluxDifference Split Scheme. 45. WrightPatterson AFB OH. 44. L. "The BaldwinLomax Turbulence Model for Two Dimensional ShockWave/BoundaryLayer Interactions. D. June 1991. R.. Vol 22. Josyula.. Kubota. Pasadena CA. July 1990. SBIR AF91108. "Numerical Solution of Threedimensional Shock Wave and Turbulent BoundaryLayer Interaction. pp431458. 40. E." Journal of Fluid Mechanics." U. NACA Technical Note 4235. pp10901096. 47." AIAA Paper 911700. Laganelli. and MacCormack. July 1984. J.the Effects of Leading Edge Bluntness and Sweep. Kittyle..38.. A. and S. FL.. M.. D. "An Experimental Study of the Interaction Between a Glancing Shock Wave and a Turbulent Boundary Layer. G. Hung. England. and Buning. Aberdeen Proving Ground.L. 196 . L. 39. and S. Observations of TurbulentBurst Geometry and Growth in Supersonic Flow. Kayser." Journal of Fluid Mechanics.
." AIAA Paper 901580. J.. R. School of Engineering. 62. Air Force Institute of Technology (AU). Number 493.. Steady State Euler Code. "Computation of Axisymmetric Base Flow with Different Turbulence Models. Investigations at Supersonic Speeds of the Base Pressure on Bodies of Revolution With and Without Sweptback Stabilizing Fins.(ed). Eglin AFB. S. and Shang. and Whitfield D. June 1991. J. Volume I . Lin. L. K." AIAA Paer 911544. Robert M. J. K. J. L. "Program EAGLE User's Manual. Moran. and R. K. WrightPatterson AFB OH. A Comparison of Molecular Vibration Modeling for Thermal Nonequilibrium Airflow. June 1955.." AGARD Conference Proceedings. Love.Multiblock Implicit. S. Washington: American Institute of Aeronautics and Astronautics. 56. MS Thesis. 57. 52. J. AFIT/GAE/ENY/90D17." AIAA Paper 911701. awaiting publication.. S. December 1990. January 1988. Loth. Cipolla. 197 . June 1990. Robert E.. Kailasanath.8. 53. D. 54. Tactical Missile Aerodynamics: Prediction Methodology. FL.. and O'Donnel." AIAA Paper 880125. Oliver." AFATLTR88117. September 1988. Volume 4. "NavierStokes Simulations of Slender Axisymmetric Shapes in Supersonic. "A Numerical Study of ThreeDimensional Separated Flows Around a Sweptback Blunt Fin. H. "An Experimental investigation of Flow Over Simple Blunt Bodies at a Nominal Mach Number of 5.. L. et. Progress in Astronautics and Aeronautics. Magagnato. Nichols." Graduate Aeronautical Laboratories California Institure of Technology (GALCIT) Hypersonics Wind Tunnel Memorandum No. E. al. and P. NACA Research Memorandum L52J21a. Lohner. 61. H.. June 1991. Beran. and P. Volume 142. D. FL. "A TwoEquation Model for Compressible Flows. M. Moran. Second Edition. "A Comparison of Molecular Vibration Modeling for Thermal Nonequilibrium Airflow. Volume IV . April 1990. Beran. Lijewski. Januaray 1990. Belk. 1992. 55. 60." AFATLTR88117. 59. Eglin AFB. "Dissipation Additons to FluxDifference Splitting." AIAA Paper 900494. Turbulent Flow. 51. Mendenhall M.Introduction and Grid Applications. McMaster. Volume 1. Mounts. Eugene S. S. F. J. R. 1952. 26.50. Moran. K." AIAA Journal. September 1988. Pasadena CA. "Program EAGLE User's Manual. 58. "Supersonic Flow Over an Axisymmetric BackwardFacing Step.
J. Thomas H.. H. 1972. "Investigation of Turbulent Separated Flows in the Vicinity of Fin Type Protuberances at Supersonic Mach Number. and W. 67. Roache.. S. Frank M. 1952. Riner. Riedelbauch.. 75." AIAA Journal. Stark. E. L. Vol. December 1993. P. 198 . Volume 22. P. June 1989. Contract Number: F0863591C0002. 65. J. "Stagnation Point HeatTransfer Measurements in Dissociatted Air. Parameter Vectors. 71. Accuracy. Rose. October.63."Journal of Computational Physics. Reller. 73. "Effective Treatments of the Singular Line Boundary Problem for Three Dimensional Grids. Computational Fluid Dynamics. Shang. NACA Research Memorandum A52E20. 2. S. and Joseph L. Price.. S. and Hamaker. 69. 64. 25. Hankey. Venkatapathy. AIAA850360. W. "Approximate Riemann Solvers. and Petty. Volume 43: 357372. A. W.98. W. Steger. G. L. S. 1981. Roe." Journal of the Aeronautical Sciences. and E." AIAA Paper 901492. Shang.. February 1958. J." Wright Laboratory Armament Directorate.. 74. Pulliam. Brenner. C. No. June 1990. John 0. An Experimental Investigation of the Base Pressure Characteristicsof Nonlifting Bodies of Revolution at Mach Numbers From 2. Private Communication. 1991. Albuquerque: Hermosa Publishers. "Numerical Simulation of Laminar Hypersonic Flow Past Blunt Bodies Including High Temperature Effects." AIAA Journal. Patrick J." NASA TN D3804.73 to 4. Palmer. February 1967.. L. 68. January 1992. J. and G. R. and Convergence for Implicit Approximate Factorization Algorithms. Volume 17. "Aerodynamic Force and Stability Predictions for the HART Missile at High Mach Numbers. October 1984. R. "Recent Improvements in Efficiency. Shang. Eglin AFB. FL. Riner. 70. "Numerical Simulation of WingFuselage Aerodynamic Interaction. Number 10: 13451353. Number 10: pp706713. "Threedimensional Supersonic Interacting Turbulent Flow Along a Corner. and Difference Schemes."AIAA 23rd Aerospace Sciences Meeting. 66. 72. Contractor Report. (January 1985). I. C. and Stallings. July 1979.. Private Communication." AIAA Paper 920545.
W. "Hypersonic BoundaryLayer Transition. Federal Republic of Germany.S. Sturek.. ICASE Report 8774. W. B.5 and 2. (1987). NACA Technical Note 2360. K. Mylin. J. F. 78. "Application of CFD to the Aerodynamics of Spinning Shell. Sturek. Speziale. 77. Stetson. January 1990. Aberdeen Proving Ground." U. Chawner. C. Sturek. W. and Nietubicz. 85. J. July 1990. A. MD. MD. ARBRLTR02501. John R..76. B. AGARD Report Number 764. J. B. C. W. "Experimental Measurements of the Turbulent Boundary Layer on a Yawed Spinning Slender Body. Army Ballistic Research Laboratory/ARRADCOM.S." U. Steinbrenner.0. G." AIAA Paper 821341. July 1975. Guidos. Z. L. Short Course in Hypersonics.. U. Army Ballistic Research Laboratory. Army Ballistic Research Laboratory. C. The GRIDGEN 3D Multiple Block Grid Generation System. Guidos.S. 80. 1951. Aberdeeb Proving Ground." AIAA Paper 840323. Effect of Tail Surfaces on the Base Drag of a Body of Revolution at Mach Numbers of 1. 86. and Dickey. Thompson. Spahr. 82. Sturek. R. B.. "Glancing ShockBoundary Layer Interactions. "NavierStokes Computational Study of the Infulence of Shell Geometry on the Magnus Effect at Supersonic Speeds. Sturek. 81. October 1990. 83. R. University of Aachen. "Preliminary Surveys of the Three Dimensional Boundary Layer on a Yawed Spinning Body of Revolution. Numerical Solution of Turbulent Flow Past a Backward Facing Step Using a Nonlinear k . J. Wright Research and Development Center.. Flight Dynamics Laboratory. MD. F. J. Aachen. WrightPatterson AFB OH. J. New York: NorthHolland. Aberdeen Proving Ground.. Fouts. and Nietubicz." Special Course on ThreeDimensional Supersonic/Hypersonic Flows Including Separation. B. BRLMR2501. 79." The Third Joint Europe/U.. 199 ." U. and Danberg. J. B.f Model. WRDCTR903022.. August 1982. and Chris L. 1985. Januaray 1984.S... and Mastin. BRLR1954. D.. C. C. R.. W. E. John P. Warsi. W. and Tuan Ngo. Stollery.. Volume 1: Final Report.. January 1977. J. B. "NavierStokes Computational Study of the Magnus Effect on Shell with Small Bluntness at Supersonic Speeds. 84. Numerical Grid Generation Foundations and Applications.. June 1983.
B. A SecondOrder Sequel to Gudonov's Method. et. M. Guidos.Surface Generation Code. Winchenbach. Eglin AFB. January 1994." AFATLTR88117.. L. and Gatlin... B. January 1972. Palo Alto CA. Y. J. 90. J.. Visbal M. 93. P. M.S. September 1988. Technical Report. S. "An Analysis of FiniteDifference and FiniteVolume Formulations of Conservation Laws. 92. 99.. December 1988. December 1988... al. al. B. A. August 1986. FL. Riner. "Program EAGLE User's Manual. "Aerodynamic Test and Ongoing Analysis of a Slender Generic Missile Configuration. Vol. "Experimental Investigation of a Fin Protuberance Paritally Immersed in a Turbulent Boundary Layer at Mach 5. AFWALTR863013. September 1988. B. OH. Volume 3. "Towards the Ultimate Conservative Difference Scheme V. R." Journal of Computational Physics. Wada. Vukelich.User's Manual. Vitale. 96. "Program EAGLE User's Manual. 1979. 97. et. 200 . D. Vukelich. Winkelmann. B. 98. "PNS Computations for Spinning and FinStabilized Projectiles at Supersonic Speeds. and Sturek.. "A Flux Splitting Scheme with HighResolution and Robustness for Discontinuities. January 1987.. E." U. Flight Dynamics Laboratory. 94." Naval Ordnance Laboratory Report NOLTR7233. TR 86508." AIAA Paer 871104. G. FL." NASA CR 177416. F. H. et al. Thompson.Grid Generation Code.Final Report. "Missile DATCOM VOL I . F.87. L. 95. F. "Missile DATCOM VOL II . 91. 89. 32... Volume III . Volume 1.." AWALTR863091. and Liou." AIAA Paper 940083. "Applying the ZEUS Code. WPAFB. Army Ballistic Research Laboratory. and Priolo. S. "A Comparison of Numerical Flux Formula for the Euler Equations and NavierStokes Equations. J." AIAA Paper 912895. Volume 2. and W. E. B. Thompson.. Wardlaw. Calculation of Viscous Transonic Flows about a Supercritical Airfoil.. Van Leer." AFATLTR88117. Volume 2. Eglin AFB. and Gatlin. C. Vinokur. Van Leer. Volume II . W.. 88. Kayser. Weinacht.. December 1986. June 1986. MD. R. Aberdeen Proving Ground. J." AWALTR863091. AFWAL. ARBRLMR3464. A." Naval Surface Warfare Center. September 1985.
Harten. R. 101. and A." Journal of Sound and Vibration. White. Inc." NASA TM 101088." Journal of Computational Physics.. C. "Analysis of Limit Cycle Flutter of an Airfoil in Incompressible Flow. Whitcomb. A Study of the ZeroLift Drag Rise Characteristicsof WingBody Combinations Near the Speed of Sound. Yee. and L. C. H. NACA Technical Report 1273. Warming.. Z. C. Moffett Field California: Ames Research Center.100. 102. R. F. 1991. Viscous Fluid Flow. Volume 57: 327360 (1985) 104. T. H. Yang. February 1989. 201 . "Implicit Total Variation Diminishing (TVD) Schemes for SteadyState Calculations. Frank M. Zhao. "A Class of HighResolution Explicit and Implicit ShockCapturing Methods. Vol 123. Yee. Second Edition. 1988. C. 1956. 103. New York: McGrawHill.
He immediately entered the Graduate School of Engineering.Vita Captain Kenneth J. he received a commission in the USAF through the ROTC Program. Permanent address: 215 Colonial Lane Dayton. Strategic Air Command. Ohio 45429 202 . he served as flight test engineer for the 49th Test Squadron. he was awarded the degree of Master of Science in Aeronautical Engineering in December 1990. Following his work at the 49th Test Squadron. Louisianna. Barksdale Air Force Base. Upon graduation. where he flew aboard many different aircraft while conducting operational 'esting of strategic weapon systems. Air Force Institute of Technology in 1989. He graduated from Robinson High School in 1980 and attended Virginia Polytechnic Institute and State University. Minnesota. After being commissioned. where he was a distinguished graduate. receiving the degree of Bachelor of Aerospace and Ocean Engineering in December 1984. Moran was born 21 April 1962 in Shaska. he attended Squadron Officer School in 1989. where as a distinguished graduate. He entered the doctoral program in January 1991.
The effects of nosetip blunting and boundarylayer condition are demonstrated. Ir +'O•tP. TITLE AND SUBTITLE Dissertation 2 Jan 91 . algorithm is implemented and employed to solve the NavierStokes equations. .' %A. Capt.C.1'a.I'r *. 5. e V7'.. turbulence.•f'li)' 4 VW'p C". Bruce Simpson Armament Directorate Wright Laboratory (WL/MNAA) AGENCY REPORT NUMBER Eglin AFB FL 32542 11. ABSTRACT (Maximurn 200 woras) sonic missile configurations can be flown at Mach numbers higher than 5. The structure lof the flow near the fins is significantly affected by the turbulent transport of momentum in regions of blocked cross flow. laminar. SPONSORING/MONITORING Dr.* 1 and tw~o 1.U/0 l0?'.~a . fluxdifferencesplitting. secondorderaccurate. Turbulence and the blockage phenomenon cause bleeding around the fin leading edges. 17 7r':?'.. REPORT DATE 3.rct (U/040180 ý.. DISTRIBUTION /AVAILABILITY STATEMENT 12b. stable beyond Math 7. NUMBER OF PAGES 224 PICKE CODE OF REP3NC 1UNCLASSIFIED !UNCLASSIFIED UNCLASSIFIED 'Unlimited ..tI..differencesplitting. The present results also suggest that the clippeddeltafin configuration is Fin Stabilized. r j %.71O ... f'5~ . Hypersonic Applied Research Technology Missile. AUTHOR(S) Kenneth John Moran. AGENCY USE ONLY (Leave blank) 2.. Stability missile aerodynamics..L 04o01 'llH++~llV%.P. The current predictions indicated that the pitchingmoment coefficient decreases with increasing Mach number much less than previous numerical computations. PERFORMING ORGANIZATION 7.REPORT DOCUMENTATION PAGE MIf. finite volume.4m~tl•O•~ () V+p S.. USAF B. REPORT TYPE AND DATES COVERED April 1994 4. FUNDING NUMBERS An Aerodynamic and StaticStability Analysis of the Hypersonic Applied Research Technology Missile 6.cIV. unlimited distribution. high fineness ratio.30 Apr 94 S. *'. SUPPLEMENTARY NOTES 12a. The aerodynamic and staticstability characteristics are investigated to determine if conventional super• 13. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) Graduate School of Engineering Air Force Institute of Technology 2750 P Street WrightPatterson AFB OH 45433 9. drag. The aerodynamic characteristics of the HART missile are predicted at Mach numbers beyond the experimental freeflight testing capabilities. The flow about the complete Hypersonic Applied Research Technology (HART) missile is simulated for inviscid.t Keat.noI t. n ttm . 1 2O _j .' The equations are solved using a finitevolume methodology.asn ýa. NavierStokes equations. Computational Fluid Dynamics.' flu3. '•0 ua'~. 14. and turbulent conditions and Mach numbers from 2 to 6. SPONSORING 'MON:7ORING AGENCY NAME(S. An explicit.fl . AND ADDRESS(ES) REPORT NUMBER AFIT/DS/AA/943 10. Ultimately.i' t.i Ap qtn . this results in lower fin effectiveness and reduced static stability. SUBJECT TERMS SECUR3T?" CL.+o " • t.n ' 17 1UOJ 1. P .rr .. DISTRIBUTION CODE Approved for public release.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.