,",_,_1/" • J
200e
c.2 v.2
COMPUTATIONAL FLUID DYNAMICS VOLUME II FOURTH EDITION
.' ··r
Klaus s: Hoffmann
Steve T.,Cbiang
. \ :  :. ,'. ,; \~. ·'~~.:!~,f:·I~·:·
.. ,::,,'.~
Fourth Edition
COMPUTATIONAL FLUID DYJ~AMICS VOLUME II
KLAUS A. HOFFMANN
STEVE T. CHIANG
ODTO KOTtJPH .. NESt1 1.,1. E. T. U. L1DRA RY
A Publication of Engineering Education System™, Wichita, Kansaa, 672081078, USA
www.EESbooks.com
Copyright @2000, 1998, 1993, 1989 by Engineering Education System. All rights reserved. No part of this publication may be reproduced or distributed in any form or by any means, mechanical or electronic, including photocopying, recording, storage or retrieval system, without prior written permission from the publisher.
The data and information published in this book are for information purposes only. The authors and publisher have used their best effort in preparing this book. The authors and publisher are not liable for any injury or damage due to use, reliance, or performance of materials appearing in this book.
ISBN 0962373133 First Print: August 2000
This book is typeset by Jeanie Duvall dba Sci'Iech Computer Typesetting of Austin, Texas.
To obtain information on purchasing this or other texts published by EES, please write to:
Engineering Education System™ P.O. Box 20078
Wichita, KS 672081078
USA
Or visit:
www.EESbooks.com
CONTENTS
Preface
Chapter Ten:
A Review
1
10.1 Introductory Remarks
10.2 Classification of Partial Differential Equations 10.2.1 Linear and Nonlinear PDEs
10.2.2 Classification Based on Characteristics 10.3 Boundary Conditions
10.4 Finite Difference Equations 10.4.1 Parabolic Equations
10.4.1.1 OneSpace Dimension 10.4.1.2 MultiSpace Dimensions 10.4.2 Elliptic Equations
10.4.3 Hyperbolic Equations
10.4.3.1 Linear Equations
10.4.3.2 Nonlinear Equations 10.5 Stability Analysis
10.6 Error Analysis
10.7 Grid Generation  Structured Grids
to.8 Transformation of the Equations from the Physical Space to Computational Space
1 1 1 2 3 4 4 4 7 8
10 10 13 14 15 16
19
ii
Contents
Chapter Eleven:
Transformation of the Equations of Fluid Motion
from Physical Space to Computational Space 21
11.1 Introductory Remarks 21
11.2 Generalized Coordinate Transformation 22
11.2.1 Equations for the Metrics 23
11.3 Nondimensionalization of the Equations of Fluid Motion 25
11.4 N avierStokes Equations 28
11.4.1 Linearization 31
11.4.2 Inviscid and Viscous Jacobian Matrices 33
11.5 ThinLayer Approximation 57
11.6 Parabolized N avierStokes Equations 60
11.7 TwoDimensional Planar or Axisymmetric Formulation 69
11.8 Incompressible N avierStokes Equations 85
11.8.1 Inviscid and Viscous Jacobian Matrices 89
11.8.2 TwoDimensional Incompressible NavierStokes Equations 91
11.9 Problems 94 Chapter Twelve:
Euler Equations 97
12.1 Introductory Remarks 97
12.2 Euler Equations 98
12.3 Quasi OneDimensional Euler Equations 99
12.3.1 Numerical Issues 100
12.3.2 Explicit Formulations 107
12.3.2.1 Steger and Warming Flux Vector Splitting 107
12.3.2.2 Van Leer Flux Vector Splitting 108
12.3.2.3 Modified RungeKutta Formulation 112
12.3.2.4 SecondOrder TVD Formulation 112
Contents iii
12.3.2.4.1 Harten Vee Upwind TVD 113
12.3.2.4.2 RoeSweby Upwind TVD 114
12.3.2.4.3 DavidYee Symmetric TVD 115
12.3.2.5 Modified RungeKutta Scheme with TVD 115
12.3.3 Implicit Formulations 115
12.3.3.1 Steger and Warming Flux Vector Splitting 116
12.4 Boundary Conditions 118
12.5 Application 1: Diverging Nozzle Configuration 121
12.5.1 Supersonic Inflow, Supersonic Outflow 124
12.5.1.1 Analytical Solution 124
12.5.2.2 Numerical Solutions 125
12.5.2 Supersonic Inflow, Subsonic Outflow 133
12.5.2.1 Analytical Solution 133
12.5.2.2 Numerical Solutions 134
12.6 Grid Clustering 144
12.7 Global Time Step and Local Time Step 146
12.8 Application 2: Shock Tube or Riemann Problem 152
12.8.1 Problem Description 152
12.8.2 Analytical Solution 153
12.8.3 Numerical Solution 157
12.9 TwoDimensional Planar and Axisymmetric Euler Equations 162
12.9.1 Numerical Considerations 163
12.9.2 Explicit Formulations 170
12.9.2.1 Steger and Warming Flux Vector Splitting 170
12.9.2.1.1 Matrix Manipulations 171
12.9.2.1.2 Existence of Zero Metrics Within the Domain 175
12.9.2.1.3 Eigenvector Matrices 177
12.9.2.2 Van Leer Flux Vector Splitting 178
12.9.2.3 Modified RungeKutta Formulation 180
12.9.2.4 SecondOrder TVD Formulation 181
12.9.2.4.1 Harten Vee Upwind TVD 181
12.9.2.4.2 Roe Sweby Upwind TVD 183
12.9.2.4.3 DavidYee Symmetric TVD 185
12.9.2.5 Modified RungeKutta Scheme with TVD 185
12.9.3 Boundary Conditions 186
lV
Contents
12.9.3.1 Body Surface 12.9.3.2 Symmetry 12.9.3.3 Inflow 12.9.3.4 Outflow
12.9.3.5 Boundary Conditions Based on Characteristics 12.9.3.5.1 Inflow Boundary
12.9.3.5.2 Outflow Boundary
12.9.3.5.3 Determination of Flow Variables 12.9.4 Implicit Formulations
12.9.4.1 Steger and Warming Flux Vector Splitting 12.9.4.1.1 Computation of the Jacobian Matrices 12.9.4.1.2 Boundary Conditions
12.10 Application: Axisymmetric/TwoDimensional Problems
12.10.1 Supersonic Channel Flow 12.10.1.1 Grid Generation 12.10.1.2 Numerical Scheme 12.10.1.3 Analytical Solution
12.10.1.4 The Physical Domain and Flow Conditions 12.10.1.5 Initial Conditions and Time Step
12.10.1.6 Results
12.10.2 Axisymmetric Blunt. Body 12.11 Concluding Remarks
12.12 Problems
186 189 190 190 191 192 192 192 193 194 197 197 200 200 201 203 204 205 206 206 210 215 216
Chapter Thirteen:
Parabolized NavierStokes Equations
218
13.1 Introductory Remarks 218
13.2 Governing Equations of Motion 221
13.3 Streamwise Pressure Gradient 223
13.4 Numerical Algorithm 225
13.5 Boundary Conditions 235
13.6 Extension to Three Dimensions 244 Contents
v
13.6.1 Numerical Algorithm 13.7 Numerical Damping Terms 13.8 Shock Fitting Procedure
13.8.1 Extension to ThreeDimensions 13.9 Application
13.10 Summary Objectives 13.11 Problems
244 249 250 257 261 264 265
Chapter Fourteen:
NavierStokes Equations
266
14.1 Introductory Remarks 266
14.2 NavierStokes Equations 267
14.3 ThinLayer NavierStokes Equations 268
14.4 Numerical Algorithms 269
14.4.1 Explicit Formulations 269
14.4.1.1 MacCormack Explicit Formulation 270
14.4.1.2 Flux Vector Splitting 272
14.4.1.3 Modified RungeKutta Scheme 275
14.4.2 Boundary Conditions 276
14.4.3 Implicit Formulations 277
14.4.3.1 Flux Vector Splitting 279
14.4.3.2 HigherOrder FluxVector Splitting 288
14.4.3.3 SecondOrder Accuracy in Time 290
14.4.3.4 L U Decomposition 291
14.5 Extension to ThreeDimensions 294
14.5.1 Explicit Flux Vector Splitting Scheme 295
14.5.2 Implicit Formulation 303
14.6 Concluding Remarks 304
14.7 Problems 306 Vl
Contents
Chapter Fifteen:
Boundary Conditions
307
15.1 Introductory Remarks 307
15.2 Classification of Schemes for Specification of Boundary Conditions 308
15.3 Category Two Boundary Conditions: Characteristics Based
Boundary Conditions 309
15.3.1 Mathematical Developments 309
15.3.2 Slip Wall Boundary Condition 321
15.3.2.1 Nonconservative (Primitive) Variables 322
15.3.2.2 Conservative Variables 323
15.3.3 NoSlip Wall Boundary Condition 323
15.3.3.1 Nonconservative (Primitive) Variables 323
15.3.2.2 Conservative Variables 325
15.3.4 Inflow /Outfiow Boundary Conditions 325
15.4 Category Three Boundary Conditions: Addition of Huffer Layer 327
15.5 Applications 328
15.5.1 Application 1: Moving Shock Wave 328
15.5.2 Application 2: Flow Over a Compression Corner 329
15.6 Concluding Remarks 334 Chapter Sixteen:
An Introduction to High Temperature Gases
336
16.1 Introductory Remarks 336
16.2 Fundamental Concepts 336
16.2.1 Real Gas and Perfect Gas 337
16.2.2 Partial Pressure 338
16.2.3 Frozen Flow 338
16.2.4 Equilibrium Flow 339
16.2.5 Nonequilibrium Flow 339 Contents
vii
16.2.6 Various Modes of Energy 16.2.7 Reaction Rates
16.2.8 FiveSpecies Model
339 341 343
16.3 Quasi OneDimensional Flow/Equilibrium Chemistry 344
16.4 Quasi OneDimensional Flow /Nonequilibrium Chemistry 345
16.4.1 Species Continuity Equation 346
16.4.2 Coupling Schemes 347
16.4.3 Numerical Procedure for the Loosely Coupled Scheme 347
16.5 Applications 350
16.5.1 Quasi OneDimensional Flow 350
16.5.2 TwoDimensional Axisymmetric Flow 352
16.6 Concluding Remarks 355
Chapter Seventeen:
Grid GenerationUnstructured Grids 356
17.1 Introductory Remarks 356
17.2 Domain Nodalization 357
17.3 Domain Triangulation 360
17.3.1 The Advancing Front Method 361
17.3.1.1 SimplyConnected Domain 361
17.3.1.2 MultiplyConnected Domain 365
17.3.2 The Delaunay Method 366
17.3.2.1 Geometrical Description 369
17.3.2.2 Outline of the Algorithm 370
15.3.2.3 An Illustrative Example 378
17.4 Concluding Remarks 382
17.6 Problems 383 viii
Contents
Chapter Eighteen:
Finite Volume Method
385
18.1 Introductory Remarks 385
18.2 General Description of the Finite Volume Method 387
18.2.1 Cell Centered Scheme 388
18.2.2 Nodal Point Scheme 389
18.3 TwoDimensional Heat Conduction Equation 390
18.3.1 Interior Triangles 391
18.3.2 Boundary Triangles 395
18.3.2.1 Dirichlet Type Boundary Condition 396
18.3.2.2 Neumann Type Boundary Condition 396
18.4 Flux Vector Splitting Scheme 402
18.4.1 Interior Triangles 403
18.4.2 Boundary Triangles 408
18.5 Concluding Remarks 410
18.6 Problems 411 Chapter Nineteen:
Finite Element Method
418
19.1 Introductory Remarks 418
19.2 Optimization Techniques 419
19.3 General Description and Development of the Finite Element Method 420
19.4 TwoDimensional Heat Conduction Equation 428
19.5 Construction of the Global Matrix 429
19.6 Boundary Conditions 432
19.7 Reduction of the HalfBandwidth of the Global Matrix 433 Contents
ix
I Appendices
Appendix G: An Introduction to Theory of Characteristics: Euler Equations 435
Appendix H: Computation of Pressure at the Body Surface 452
Appendix I: Rate of Formation of Species 458
References
461
Index
PREFACE
The fundamental concepts of computational schemes established in the first volume are extended to the solution of Euler equations, Parabolized NavierStokes equations, and N avierStokes equations, along with treatment of boundary conditions. In addition, chemically reacting flows, unstructured grids, finite volume schemes, and finite element method at the introductory level are included.
This volume begins with a review of the basic concepts which is presented in Chapter 10. Subsequently, the transformation of the equations of fluid motion from physical space to computational space is provided in Chapter 11. This chapter also includes the linearization of the equations as well as the derivation of the Jacobian matrices. Chapter 12 presents numerical schemes for the solution of the Euler equations for inviscid flowfields. Specifications of the boundary conditions, along with illustrated examples, are provided in this chapter. Chapter 13 presents Parabolized NavierStokes (PNS) equations and a numerical algorithm for solution. The shock fitting procedure is discussed in this chapter as well. The NavierStokes equations and various numerical schemes for solutions are discussed in Chapter 14. Specification of boundary conditions, derivation of governing equations, and comparison of several types of boundary conditions are provided in Chapter 15. An extension of the governing equations to include the effect of chemistry for hypersonic flowfield computations is included in Chapter 16. To familiarize the reader with unstructured grids which are used in conjunction with finite volume and finite element schemes, they are introduced in Chapter 17. It develops some fundamental concepts and explores two techniques for generation of unstructured grids in twodimensions. Finally, finite volume schemes and finite element method are developed at the introductory level in Chapters 18 and 19, respectively.
Several computer codes are developed based on the materials presented in this text. These codes, manuals, and additional examples are presented in the text, Student Guide to CF D Volume II.
Finally, our sincere thanks and appreciation are extended to all individuals acknowledged in the preface of the first volume. Thank you all very much for your friendship and encouragement.
Klaus A. Hoffmann Steve T. Chiang
Chapter 10
A Review
10.1 Introductory Remarks
The fundamental concepts of computational fluid dynamics were introduced in the previous chapters. Various aspects of numerical schemes were explored with regard to simple partial differential equations. In all cases up to Chapter 8, the investigations were limited to a single equation. In the upcoming chapters the concepts are extended to systems of equations. Before proceeding further, however, it is beneficial to review and summarize the content of the previous chapters.
10.2 Classification of Partial Differential Equations
Partial differential equations (PDEs) can be classified into different categories, where within each category they may be classified further into subcategories. The numerical procedure used to solve a partial differential equation very much depends on the classification of the governing equation. A brief review of the classification of partial differential equations is provided in the following subsections.
10.2.1 Linear and Nonlinear PDEs
(a) Linear PDE: There is no product of the dependent variable and/or product of its derivatives within the equation.
(b) Nonlinear PDE: The equation contains a product of the dependent variable and/or a product of the derivatives.
2
Chapter 10
10.2.2 Classification Based on Characteristics
(I) Firstorder PDE: Almost all firstorder PDEs have real characteristics, and therefore behave much like hyperbolic equations of second order.
(II) Secondorder PDE: A secondorder PDE in two independent variables, x and y, may be expressed. in a general form as
(101)
The equation is classified according to the expression (B2  4AC) as follows:
{ < 0 + elliptic equation (B2  4AC) = 0 + parabolic equation > 0 + hyperbolic equation
The following criteria may be stated with regard to each category defined above:
(a) Elliptic equations
• No real characteristic lines exist
• A disturbance propagates in all directions
• Domain of solution is a closed region
• Boundary conditions must be specified on the boundaries of the domain
(b) Parabolic equations
• Only one characteristic line exists
• A disturbance propagates along the characteristic line
• Domain of solution is an open region
• An initial condition and two boundary conditions are required
(c) Hyperbolic equations
• Two characteristic lines exist
• A disturbance propagates along the charact.eristic lines
• Domain of solution is an open region
• Two initial conditions along with two boundary conditions are required
(III) System of FirstOrder PDEs
A system of firstorder PDEs may be expressed in a vector form as
l» 8~ l»
at + [A] ax + [B] 8y + W" = 0
(102)
A Review
3
where the vector ~ contains the dependent variables. The system is classified according to the eigenvalues of coefficient matrices [A] and [B]. If the eigenvalues of matrix [A] are all real and distinct, the system is classified as hyperbolic in t and x. If the eigenvalues of [A] are complex, the system is elliptic in t and x. Similarly, the system is classified with respect to the independent variables t and y based on the eigenvalues of matrix [B].
For a steady equivalent of (102), given by
8q, 8lp
[A]  + (B] +w = 0
Bx By
(103)
the classification is as follows:
{ < 0 + elliptic
H = 0 + parabolic > 0 + hyperbolic
where
and
P=jAj, Q=IBI
For a system composed of two equations, R is given by
where
[A] = [~: ~ land [B] = [: ~
(IV) System of SecondOrder PDEs
The classification of a system of secondorder PDEs is facilitated if the secondorder PDEs are reduced to their equivalent firstorder PDEs. Subsequently, the system is classified as previously seen. The procedure could be cumbersome. For specific details and examples, Section 1.9 should be reviewed.
10.3 Boundary Conditions
A set of specific information with regard to the dependent variable and/or its derivative must be specified along the boundaries of the domain. This set of information is known as the boundary condition and may be categorized as follows.
4
Chapter 10
(a) The Dirichlet boundary condition: The value of the dependent variable along the boundary is specified.
(b) The Neumann boundary condition: The normal gradient of the dependent variable along the boundary is specified.
(c) The Mixed boundary condition: A combination of the Dirichlet and the Neumann type boundary conditions is specified.
10.4 Finite Difference Equations
The partial derivatives appearing in the differential equations are replaced by approximate algebraic expressions to provide an equivalent algebraic equation known as the finite difference equation. Subsequently, the finite difference equation is solved within a domain which has been discretized into equally spaced grids. Finite difference equations commonly used for the solution of parabolic, elliptic, and hyperbolic equations are reviewed in this section.
10.4.1 Parabolic Equations
Various finite difference formulations are reviewed for the onedimensional parabolic equations initially and, subsequently, extended to multidimensional problerns.
10.4.1.1 OneSpace Dimension
The simple diffusion equation is used in this section to review various finite difference equations. The model equation is given by
where a is assumed to be a constant and hence a linear equation. To facilitate the review process, various aspects of each finite difference formulation such as the order of accuracy, amplification factor, stability requirement, and the corresponding modified equation are summarized. In the formulations to follow I the diffusion number is designated by d, which is defined by
A Review
5
Scheme:
Formulation:
Order:
Amplification Factor:
Stability Requirement:
Modified Equation:
Special Considerations:
FTCS explicit
n+ 1 _ n + d (n 2 n + n )
Ui  U. uHl  Ui u'_1
o [(t:J.t), (t:J.x) 2]
G = 1 + 2d( cos 0  1) 1
d<
2
{Ju = a{J2u + [!a2(t:J.t) + _!_a(ll.x)2] a'u
1Jf {Jx2 2 12 {Jx·
[1 3 )2a'U 1 2( 2
+ '3a (t:J.t 8x'  12a At) (t:J.x)
1 A )2] 86u
+ 360 a(ux {Jx5 + ...
None
Scheme:
Formulation:
Order:
Amplification Factor:
Stability Requirement:
Modified Equation:
Special Considerations:
DuFortFrankel explicit
n+l _ 1  2d nl 2d (n n )
U.  1 + 2dui + 1 + 2d ui+! + Uil
o [(dt)', (dx)', (~) l
G = 1: 2d [2d cosO± (1 4cfSin20)i] Unconditionally stable
8u = afJ2u + [_!_a(t:J.x2) _ ci (t:J.t) 2] a'u Ox 8x2 12 Ax 8x4
+ [~a3(At)2 + 3!Oa(Ax)"
5 (ll.t )4]lIu
+20 ax 8x6 + ...
Requires two sets of data to proceed
6
Chapter 10
Scheme:
Formulation:
Order:
Amplification Factor:
Stability Requirement:
Modified Equation:
Special Considerations:
Laasonen implicit
dui!l  (1 + 2d)ui+ 1 + d1Li_l = U':
o [(L\t) , (L\X)2]
G= 1
1 + 2d(1  CosO)
Unconditionally stable
8u 82u [1 2( ) 1 )2] ffu 1ft = 0 8x2 + '20 L\t I 120(L\x 8x4
[1 1
+ '30?(Llt)2 + 1202(At) (ax)2
1 ( )4] flu
+ 3600 ax 8x6 + ...
Requires solution of a tridiagonal system
Scheme:
Formulation:
Order:
Amplification Factor:
Stability Requirement:
Modified Equation:
Special Considerations:
CrankNicolson implicit
1 d n+ 1 ( d) n+ 1 1. d n+ lIn
'2 UHl  1 + U, + 2 U,l = '2uH1
+ (d  1)u~  !du~ 1
s 2 s
O [(At)2, (L\X)2]
G = 1  d( 1  CosB)
1 + d(1  CosB)
Unconditionally stable
8u = oflu + [_!_0(AX)2] B'u 1Jx 8X2 12 8x'
+ [:203(At)2 + 3!00(AX)4] fxv + ... Requires solution of a tridiagonal system
A Review
10.4.1.2 MultiSpace Dimensions
The review of multidimensional problems will be limited to twospace dimensions. The procedure to threespace dimensions is similar. However, it is cautioned that the extension may not be trivial, and certain formulations which may have been unconditionally stable in twospace dimensions may become only conditionally stable in threespace dimensions. The model equation used is the diffusion equation in twaspace dimensions given by
au = (82u B2u) 8t Q 8x2 + 8y2
Scheme:
FTCS explicit
Formulation:
U'11 = ui.; + dz(u~+l';  2ui.; + U~l';) + d" (U~';+l 2u':'· + u~· 1)
'.:I '.:1
Order:
1 Stability Requirement: (dz + c4) ~ '2
Scheme:
ADI
Formulation:
and
 (~d,) U~J21 + (1 + d,)U~l  (~d,) U~~l = (~dz) U~:lt.; + (1  dz)u:;l + (~dz) ~~1~ o [(At)2, (AX)2, (Ay)2]
G = [1  dz(l  CosOz)1 [1  d,(1  CosO,)1 [1 + lIz(1  CosOz)] [1 + dy(1  CosO,)] Unconditionally stable
Order:
Amplification Factor:
Stability Requirement:
7
8
Chapter 10
Scheme:
Fractional step
Formulation:
n+1 n+l n+1
~Ui+l~  (1 + 2dz}u'J + dZU,_l'J,j
= dzu~+lJ + (2d:z;  1}1I,'.;  dzU~lJ
and
Order:
o [(~t)2, (~X)2, (~y):~] Unconditionally stable
Stability Requirement:
10.4.2 Elliptic Equations
The model equation utilized to represent various finite difference equations is the Poisson equation expressed in twospace dimensions given by
Only iterative schemes which are usually the most efficient schemes to solve elliptic equations are reviewed in this section. In the formulations to follow, the ratio of
. . desi ed b f3' f3 ~x
stepsizes 18 esignat y, i.e., = fly'
I Scheme: Point GaussSeidel (PGS)
Formulation:
Order:
Scheme:
Line GaussSeidel (LGS)
Formulation: (x direction)
Order:
ffu lfu _ 1 (A _)2 &4u 1 (A )2 tru Modified Equation: 8x2 + 8y2   12 LXI; 8x'  12 uy 8y4 + ...
A Review
Scheme:
Point Successive OverRelaxation (PSOR)
Formulation:
Special Considerations: The range of relaxation parameter is 1 ::; W < 2
Scheme: Line Successive OverRelaxation (LSOR)
Formulation:
(x direction) wUf!lJ  2 (1 + fJ2) Ufjl + wu.~:l,;
= (1  w) [2(1 + ,(2)] U~J  w,82 (U~';+1 + U~~l) Special Considerations: The range of relaxation parameter is 1 < w < 2
Scheme:
AD!
Formulation:
Hl ( 2) Hi k+i _ 2 ( k k+i )
U'l,;  2 1 + fJ Ui,; + Ui+l,j  f3 Ui,;+l + Ui,jl
and
/.12 J::+l 2(1 /l2) k+1 + fl2 k+l Hi HI
fJ Ui,jl  + fJ Ui,; fJ Ui,j+l = UHI,j  Uil,j
Scheme:
ADI with relaxation parameter
k+l 2(1 fl2) k+i k+l
WUil,j  + fJ Ui,; + WUi+l,;
Formulation:
= (1  w) [2 (1 + {P)] u~,;  W {j2 ( U~,;+l + U:;.!l)
and W {j2u~j!.1  2(1 + ,B2)~t + W ,B2~:Ul
=  (1  w) [2 (1 + fJ2)] u~;l  w (U~:l~ + uf!l,; )
9
10
Chapter 10
10.4.3 Hyperbolic Equations
Investigation of various finite difference equations is easily accomplished with regard to linear hyperbolic equations. Subsequently, the conclusions may be extended to nonlinear hyperbolic equations. With that in mind, the review of the formulations is performed sequentially in two parts.
10.4.3.1 Linear Equations
The wave equation given by au au
at = a ax' a > 0
is used to review linear hyperbolic equations. Note that the speed of sound, a, in the equation above is assumed to be a constant and, hence, a linear equation. The parameter, at:J..t/ t:u, defined as the Courant number and designated bye, will be used in the formulations to follow.
Scheme:
Formulation:
First upwind differencing u~+1 = uf  e(ur  U;'_l)
Order:
o [(Llt), (Llx)]
G = 1 e(l  CosfJ)  i(cSinfJ)
Amplification Factor:
Stability Requirement:
Modified Equation:
Scheme:
Lax
n+l_I(n n) l(n n)
u, "2 Ui+l + Uil  '2e Ui+l  Uil
o [CAt), (~;) 1
G = Cos fJ  i(eSin 8) e$1
au au 1 ( (I ) (fu 7ft = aFx + '2a Ax) c  e 8x2
1 2 lfiu
+ aa(Llx) (1  c) ax"! + ...
Formulation:
Order:
Amplification Factor:
Stability Requirement:
Modified Equation:
A Review
11
Scheme:
Formulation:
Order:
Amplification Factor:
Stability Requirement:
Modified Equation:
Special Considerations:
Midpoint Leapfrog
Un+1  Un1 C (un Un)
i  iHI  iI
o [(6.t)2, (6.X)2 J
G = ± [1 c2Sin20]i  i(cSinO)
c:51
& au 1 ( )2( ~)&u
7Jf = alJX  "6a ax 1  c: 8x3 + ...
Requires two sets of data for the solution
to proceed
Scheme:
Formulation:
Order:
Amplification Factor:
Stability Requirement:
Modified Equation:
Lax Wendroff
n+1 n l(n n)
u· = u·  c U'+l  u· 1
, '2' ,
+ ~c? (U~l  2uf + Ui'l) o [(6.t)2, (AX)2]
G = 1  c? (1  CosO)  i(c Sin 0)
12
Chapter 10
Scheme:
Formulation:
Order:
Amplification Factor:
Stability Requirement:
Modified Equation:
BTCS implicit
o [(~t) I (~X)2] G = 1  i(cSinO) 1 + c2 (Sin2 0 )
None
Scheme:
Formulation:
Order:
Amplification Factor:
Stability Requirement:
Modified Equation:
CrankNicolson
!CU~++11  U~+l  !CU~+ll = u~  !C(U~+l  U~ 1)
4' , 4 , '4 1 1
o [(At)2 , (Ax)2]
G = 1 O.Si(cSinlJ) 1 + O.Si (cSinlJ)
None
Scheme:
Formulation:
and
Order:
Stability Requirement:
Lax Wendroff
1 ( n+l n+l)
u~+ = u~  c U. J  U. J
, , ttl '1
c<l
A Review
Scheme:
MacCormack
Formulation:
and
n+1 1 [ n +. (. .)] u· =  u· u·  c u·  u·
1 2 1 , ,.
Order:
Stability Requirement:
10.4.3.2 Nonlinear Equations
The inviscid Burgers equation is used to review various schemes for the solution of hyperbolic equations. Recall that the model equation is given by
81.1. 81.1. 8E
=1.1.=
8t 8x 8x
where E = ~U2. Now, the Courant number is defined as c = u6.t/6.x.
Scheme: Lax
I • n+l _ 1 ( n n) 1 6.t (DR DR)
Formu attorn 1.1.,  '2 Ui+l + U,_1  '2 6.x £.<H1  £.<,_1
Order: 0 [(6.t) , (6.x) 2]
Amplification Factor: G = CosO  i(c Sin 0)
Stability Requirement: jumax !! I s 1
Scheme: Lax Wendroff
n+l _ n 1 6.t (~ 1':'II'l) 1 (6.t) 2
Formulation: u,  Ui  '2 Ax £.<i+l  £'<i1 + 4 Ax
[(u~+1 + u~) (Er+l  Er)  (u~ + U~l) (Er  Ei'1)] Amplification Factor: G = 1 2c2(1 CosO)  2i(cSinO)
Stability Requirement: lumax ~! I :5 1
13
14
Chapter 10
Scheme:
MacCormack
and
U: = ui  ~! (E;+1  E;)
U~+1 = ~ [u'! + u~  .t1t (e  E~ 1)J
' 2' , .t1x' ,
o [(.t1t)2 , (.t1x) 2]
Formulation:
Order:
10.5 Stability Analysis
The error introduced in the finite difference equations due to the truncation of the higher order terms in the Taylor series expansion may grow unbounded, producing an unstable solution. The control of errors within the solution is of primary concern for any numerical scheme. To establish the necessary requirements, a stability analysis must be performed. Among various methods available for stability analysis are: (1) The discrete perturbation stability analysis, (2) The von Neumann (or Fourier) stability analysis, and (3) The matrix method. It should be noted that direct stability analysis of a nonlinear, multidimensional, coupled system of equations is usually cumbersome. In most cases, expressions are proposed which are based on the stability analysis of simple model equations complimented and reinforced by numerical experimentation. Thus, one encounters the suggested stability requirement for a particular scheme which resembles those of simple model equations, but includes some modifications based on numerical experimentations.
To review the limitations and conclusions provided from the von Neumann stability analysis, the summary stated in Chapter 4 is repeated at this point.
1. The von Neumann stability analysis can be applied to linear equations only.
2. The influence of the boundary conditions on the stability of the solution is not included.
3. For a scalar PDE which is approximated by a twolevel FDE, the mathematical requirement is imposed on the amplification factor G as follows:
(a) if G is real, then IGI ~ 1
(b) if G is complex, then IGI2 < 1, where IGI2 = GO
4. For a scalar PDE which is approximated by a threelevel FDE, the amplification factor is a matrix. In this case, the requirement is imposed on the eigenvalues of G as follows:
A Review
15
(a) if A is real, then IAI < 1
(b) if A is complex, then IAI2 :s 1
5. The method can be easily extended to multidimensional problems.
6. The procedure can be used for stability analysis of a system of linear PDEs.
The requirement is imposed on the largest eigenvalue of the amplification matrix.
7. Benchmark values for the stability of unsteady onedimensional problems may be established as follows:
(a) For most explicit formulations:
I. Courant number, c s 1
II. Diffusion number, d < .!
2
III. Cell Reynolds number, Ree::; (2/c)
(b) For implicit formulation, most are unconditionally stable.
8. For multidimensional problems with equal grid spacing in all spatial directions, the stated benchmark values are adjusted usually by dividing them by the number of spatial dimensions.
9. On occasions where the amplification factor is a difficult expression to analyze, graphical solution along with some numerical experimentation will facilitate the analysis.
10.6 Error Analysis
The truncation of terms in the approximation of a partial derivative could begin from an oddorder or an evenorder derivative term. For example, one may approximate a firstorder derivative by either
(104)
or
au 11,,+1  Uil (6.X)2 (J3u
ax = 26.x + 3t ax! + ...
The approximation (104) may be truncated and expressed as
au = Ui+l  11" + O(6.x)
ax 6.x
(105)
16
Chapter 10
where the dominant (or leading) error term includes a secondorder derivative, i.e., even. The second expression given by (105) is written as
=: = u'+~~:l + O(~X)2
where the dominant error term now includes an odd derivative. The behavior of error associated with finite difference equations is strongly influenced by the dominant error term. To clarify the types of error introduced to the finite difference equations, a convective dominated equation, where physical viscosity is absent, will be used. Thus, consider the wave equation and two different finite difference equations given by
U~+1 = u~  c(u~  u'! 1)
1 1 1 1
(106)
and
Un+1  un1 C(un un)
,  iHI  iI
(107)
The FDEs are recognized as the first upwind differencing scheme and the midpoint leapfrog method. To identify the dominant error term of an FDE, the modified equation must be investigated. The corresponding modified equations for the FD Es given by (106) and (107) are, respectively:
au au 1 [J2u 1 2 ( ) [J3u
at = a ax + 20(&;)(1 c) ax'  So(&;) 2e Scl 1 ax3 + ...
(108)
and
(109)
Observe that the dominant error terms in Equations (108) and (109) include secondorder derivative and thirdorder derivative, respectively. Recall that, from a physical point of view, a secondorder derivative is associated with diffusion. Indeed, the coefficient of the secondorder derivative in Equation (108) is known as the numerical viscosity. Thus, it is obvious that the error associated with Equation (108) is dissipative and, hence, it is called dissipation error, On the other hand, an FDE scheme, where its corresponding modified equation possesses an oddorder derivative as the lead term in error, is associated with oscUlations within the solution. Such an error is called dispersion error,
10.7 Grid GenerationStructured Grids
Finite difference equations are most efficiently solved in a rectangular domain (for 2D applications and an equivalent hexahedral domain for 3D applications)
A Review
17
with equal grid spacings. Unfortunately, the majority of physical domains encountered are nonrectangular in shape. Thus, it is necessary to transform the nonrectangular physical domain to a rectangular computational domain where grid points are distributed at equal spacings. It is also important to note that the transformation allows the alignment of one of the coordinates along the body, thus facilitating the implementation of the boundary conditions. The objective of grid generation is then to identify the location of the grid points in the computational domain and the location of the corresponding grid points in the physical space. Furthermore, the metrics and Jacobian of transformation which are required. for the solution of flow equations are computed within the grid generation routine.
Typically, grid generation schemes may be categorized as algebraic methods or differential methods. In the latter case, the scheme is based on the solution of a set of PDEs and may be subcategorized as either an elliptic, parabolic, or hyperbolic grid generation. Either category of grid generation scheme should include the following considerations.
1. A mapping which guarantees onetoone correspondence ensuring grid lines of
the same family do not cross each other;
2. Smoothness of the grid distribution;
3. Orthogonality or near orthogonality of the grid lines;
4. Options for grid clustering.
A brief summary of the advantages and disadvantages of each method is provided below.
1. Algebraic grids
The ad vantages of this category of grid generators are:
(a) They are very fast computationally;
(b) Metrics may be evaluated analytically, thus avoiding numerical errors; (c) The ability to cluster grid points in different regions can be easily imple
mented.
The disadvantages are:
(a) Discontinuities at a boundary may propagate into the interior region which could lead to errors due to sudden changes in the metrics;
(b) Smoothness and skewness may be difficult to control.
2. Elliptic grids
The advantages of this class of grid generators are:
18
Chapter 10
(a) Will provide smooth grid point distribution, i.e., if a boundary discontinuity point exists, it will be smoothed out in the interior domain;
(b) Numerous options for grid clustering and surface orthogonality are available;
(c) Method can be extended to 3D problems.
The disadvantages of the method are:
(a) Computation time is large (compared to algebraic methods or hyperbolic grid generators);
(b) Specification of the forcing functions P and Q (or the constants used in these functions) is not easy;
(c) Metrics must be computed numerically.
3. Hyperbolic grids
The advantages of hyperbolic grid generators are:
(a) The grid system is orthogonal in two dimensions;
(b) Since a marching scheme is used for the solution of the system, computationally they are much faster compared to elliptic systems;
(c) Grid line spacing may be controlled by the cell area or arclength functions.
The disadvantages are:
(a) Boundary discontinuity may be propagated into the interior domain; (b) Specifying the cellarea or arclength functions must be handled carefully.
A bad selection of these functions easily leads to undesirable grid systems.
Finally, the metrics and Jacobian of tranformation are given by the following expressions.
1. Two dimensions
& = Jy" ~y = JX'1 T}:z: = JYe "hi = JXe
where
1 J=
xeY"  Yex"
A Review
19
2. Three dimensions
ez = J(YfJz(  Y(ZfJ) ell = J(x(z"  xfJzt;} ~z = J(xfJY(  x(YJ]) TJz = J(y,ze  y{zc) T/y = J(x{z(  x,ze) TJ% = J(XCYe  xeyt;} ~ = J(Yez"  YfJze) (If = J(x'1ze  xez,,) (z  J(xeYfJ  x'1Ye)
e, =  (xr& + Vrell + z.,.ez) 1Jt = (xrTJz + YrTJ,I + ZrTJz) (, =  (xr<"z + Yr(lI + z.,.(z)
where
J = 8(e, 11, () = 1
a(x, y, z) xe(Yflze  YCz,,)  x'I(Yeze  Veze) + xc(Yez"  Yflze)
10.8 Transformation of the Equations From the Physical Space to Computational Space
The partial derivatives expressed in the physical space are related to the partial
derivatives in the computational space by the following relations:
8t = Ir + €tk + 1Jt~ + (t&
Ix = ez& + TJzlr, + t:zk
By = ell& + TJ'J6r, + (,1&
Ix = ez& + 77z8q + (%&
The NavierStokes equations in a flux vector form may be expressed in the physical space by
8Q BE er 8G es; 8Fv eo, at + 8x + 8y + 8z = 8x + By + 8z
can be transformed to the computational space and expressed by 8Q 8E 8F 8G 8Ev 8Fv 8Gv 8T + 8e + 8TJ + 8t; = 8e + 8TJ + 8t;
20
Chapter 10
where
 Q Q= J
E = ~(~tQ + ~l:E + ~yF + ~zG)  1
F = J(f}tQ + 7Jl:E + 7JIIF + 7JzG)
 1
G = J«tQ + (l:E + (yF + (zG)
 1
Ev = J(~%Ev + {IIFv + {zGv}
 1
r, = J(7Jl:Ev + 'lIlFv + 7JzGv)
 1
Gu = J«(l:Ev + (IIFv + (zGv)
The inviscid and viscous Jacobian matrices which are produced in the process of linearization of the equations are given in Chapter 11 for the N avierStokes, ThinLayer NavierStokes, Euler, and Parabolized NavierStokes equations.
Chapter 11
Transformation of the Equations
of Fluid Motion from Physical Space to Computational Space
11.1 Introductory Remarks
To enhance the efficiency and accuracy of a numerical scheme and to simplify implementation of boundary conditions, a transformation from physical space to computational space is performed. This transformation allows clustering of grid points in regions where flow variables undergo high gradients and grid point motion when required. The computational domain is a rectangular shape which is divided into an equally spaced grid system. In order to solve the governing equations of motion in the computational space, a transformation of the equations from physical space into computational space is required. Any assumption on the simplification of the equations of motion is imposed on the transformed equations. For example, to reduce computational time and required storage, the full NavierStokes equations may be simplified by neglecting the circumferential and streamwise gradient of stresses, while retaining only the normal gradient of the stresses. The resulting equations are known as the ThinLayer NavierStokes equations. The reduction of equations is performed on the transformed equations.
This chapter investigates generalized coordinate transformation of the governing equations of fluid motion expressed in the Cartesian coordinate system (x, y, z) from physical space to computational space (~, 1], (). Various formulations of the equations which are discussed include full N avierStokes (NS) I Euler, ThinLayer NavierStokes (TLNS), and Parabolized NavierStokes (PNS) equations. A summary of the assumptions which are imposed in the reduction process of the
22
Chapter 11
equations is presented. The formulations include threedimensional flows as well as twodimensional planar or axisymmetric flow fields.
In order to include the capability for shock capturing, the equations of motion are written in conservative form. Any set of equations is then approximated by finite difference formulation, which is solved in the rectangular grid system. Furthermore, in the linearization process Jacobian matrices are produced, which are included in this chapter.
This chapter summarizes the equations of fluid motion in computational space and presents them in concise and orderly manner. Thus, the objectives of this chapter are summarized as follows:
(1) Define the metrics and the Jacobian of transformation;
(2) Express the equations of fluid motion in a generalized coordinate system for NS, TLNS, Euler, and PNS equations; and
(3) Derive the Jacobian matrices for each set of equations which are used in various numerical algorithms.
11.2 Generalized Coordinate Transformation
The equations of motion are transformed from the physical space (x, y, z) to the computational space (e, TI, () by the following relations:
(  (t,x,y,z)
(111) (112) (113) (114)
T  t
e  e(t,x,y,z)
TI  .,.,(t, x, y, z)
The chain rule of partial differentiation provides the following expressions for the Cartesian derivatives:
a a a a 8 (115)
at  aT + et a€ + Tft a.,., + (t 8e
8 a a a (116)
ax  e% 8€ + .,.,% 87] + (% 8(
8 8 8 8 (117)
8y  €" 8€ + 71y a7] + (" 8(
8 8 a a (118)
8z  €z ae + ""z 811 + (z 8( Transformation of the Equations of Fluid Motion
23
11.2.1 Equations for the Metrics
From Equations (115) through (118), it is obvious that the value of the metrics ~t, TIt, (t, ~r, 1}:z:, (.:r, ~y, 1}y, (y, ~Ie' TIle, and (Ie must be provided in some fashion. In most cases the analytical determination of the metrics is not possible and, therefore, they must be computed numerically. Since the stepsizes in the computational domain are equally spaced, X6 XI'I' x" etc., can be computed by various finite difference approximations. Thus, if the metrics appearing in Equations (115) through (118) can be expressed in terms of these derivatives, the numerical computation of metrics is completed. To obtain such relations, the following differential expressions are considered:
8t 8t 8t 8t
dt =  dr +  de +  d1} +  d(
8r 8~ 877 8(
But according to (111), 8t
 1 8r
8t af.
and
at aTl
8t
 0 a(
thus
dt  dr
(119)
Similarly,
dx  xTdr + x~de + x'ld1} + x,d( dy  yTdr + y~de + Y'ld77 + y,d( dz  zTdr + zed{ + Zr,dTJ + z(d(
Equations (119) through (1112) are expressed in a matrix form as
(1110) (1111) (1112)
[ : ] = [~ :: :: :~] [~dr(TJ]
dz ZT ze Zr, z,
Reversing the role of the independent variables,
(1113)
dr
dt
(1114) (1115) (1116) (1117)
df.  f.tdt + €:r:dx + €ydy + €zdz d77  Tltdt + fJrdx + f}ydy + Tlzdz d(  (tdt + (rdx + (ydy + (zdz
24 Chapter 11
which are expressed as
[~]~[l~ 0 0 ~: ][ £ ]
& ell (1118)
7]x 7]11
c, (II (1l dz
Comparing Equations (1113) and (1118), one concludes that
[ l~ 0 0 ~ ] ~ [ ~, 0 0 or
ex ell xE x" x,
7]% 1]y 7]z Yr Ye Y" l/(
(I (x (II (z z; ze z" z(
From which,
e%  J(y"z(  y(z,,) (1119)
ell  J(x(Zr,  X'1Z,) (1120)
ez  J(X'1Y(  xcV,,) (1121)
7]%  J(y,ze  Yeze) (1122)
1]y  J(xez(  x(ze) (1123)
7]z  J(x(Ye  xeYd (1124)
ex  J(Yez"  v"ze) (1125)
(II  J(x"ze  xez'1) (1126)
(z  J(xeY'1  x"Ye) (1127)
€t  (xr& + Yrell + zrez) (1128)
7]t  (xr7]% + Yr7]y + Zr7]z) (1129)
(I  (xrCx + Yr(y + Zr(z) (1130) After substitution of Equations (1119) through (1127) into Equations (1128), (1129), and (1130),
et  J[xr(Y(z"  Y'1zd + Yr(x"z(  x(z,,) + zr(x,y"  x'1Vd1 (1131)
T/t  J[Xr(Yez,  v,ze) + Yr(X(Ze  xezt;} + Zr(xeY(  X(Ye)] (1132)
(t  J[xr(V'1Ze  Yez,,) + Yr(xez"  X"zE) + Zr(X"Ve  xeV'1)] (1133)
Transformation of the Equations of Fluid Motion
25
where J is the Jacobian of transformation defined by
J = a(~, 17, () _ 1
a(x, y, z) xC(Y'f/Z(  Y(~)  x'f/(YeZ(  Y(Ze) + xdYe~  Y'f/ze)
(1134)
11.3 Nondimensionalization of the Equations of Fluid Motion
Equations of fluid motion may be nondimensionalized to achieve certain objectives. For one, it would provide conditions upon which dynamic and energetic similarity may be obtained for geometrically similar situations. Second, the solution of such equations would usually provide values within limits between zero and one. Generally a characteristic dimension, such as the chord of an airfoil or the length of a vehicle, is selected to nondimensionalize the independent spatial variables. Freestream conditions are used to nondimensionalize the dependent variables. Among many choices available, the following will be used in this and in the subsequent chapters:
e tuoo • x
 L x =
L
/t. /t • u
 u =
J.Loo Uoo Y· = I '
• z z =
L
• v v =Uoo
* w w =
Uco
p. = P T* = T p. = P 2
Poo T co Poouoo
The nondimensional parameters are defined as:
• et
e =t u2 00
Reynolds number:
Reoo = Poo uooL /too J.LCp Pr=
k
Prandtl number:
Using the nondimensional variables defined above, the equations of fluid motion in the Cartesian coordinate system are expressed as:
1. Continuity:
(1135)
2. Xcomponent of the momentum equation:
a a 2 a a
 (p·u·) +  (p·u· + p.) +  (p'u'v') +  (lu·w·) =
8t ax· ay· az·
(1136)
26
Chapter 11
For a Newtonian fluid,
• • _1_ [1t.(8U' + av'.)]
T:rll  Ty:r 
Reoo ay· ax·
• • 1 [ aw' au']
T:rz  rz:r  Reoo 11: (ox. + 0 Z'~ ) 3. Vcomponent of the momentum equation:
where
and T~ is given by (1138).
4. Zcomponent of the momentum equation:
{) (p·w·) + a (p·u·w·) + ~ (p.v.w.) +..!._ (p·w·2 + p.) =
Ot· ax· Oy· a z·
a~. (r;z) + a~. (T;z) + a~. (7·:z)
where
·  1 [2 .aw· \. n· v~·l
T  11.+", v'
%Z Reoo r: oz.
and r;% and r;% are given by Equations (1139) and (1142), respectively.
(1137)
(1138)
(1139)
(1140)
(1141)
(1142)
(1143)
(1144)
Transformation of the Equations of Fluid Motion
27
5. Energy:
(1145)
where
q! J.L. aT· (1146)
 ReooPr b  1) M! 8x·
q; /1: aT· (1147)
 ReooPr b  1) M!, 8y·
q; /1" m: (1148)
 ReooPr b  1) M! 8z· The nondimensional equations of fluid motion may be expressed in a flux vector
form as:
8Q· 8E· 8p· 8a· 8E 8F· 8a· __ + __ + __ + __ = __ v + __ v + __ v
8t· ox* oy· Bz' ox· oy· oz·
(1149)
where
p. p·u· Q. = p·v· p·w· pte;
(1150)
p·u· 2
p·u· + p.
E· = p·u·v· p·u·w*
(p* e~ + p·)u·
o
(1152)
(II51)
p·v· 0
p·v·u· •
TJI%
p.= 2 (1153) F*= r.* (1154)
p·v· + p. 11 YfI
p·v·w· •
TlIZ
(p·e~ + p·)v· u*r.· + v·r.* + w·r.* _ q*
7/% 1171 liZ 11 28
Chapter 11
C·=
p·W· p·W·U· p·W·V·
2
p·W· + p.
(p·e; + p*)W·
o
(1155)
C·= "
T· =
(1156)
T· Zl/
t;
U·T· + V·T· + W·T·  q*
Ii!% ZI/ Z% %
The notation "." which is used to designate nondimensional quantities will be dropped for the remainder of this chapter and those following. Thus, all the equations will be in nondimensional form unless otherwise specified.
11.4 NavierStokes Equations
The equations of fluid motion in complete form which include the conservation of mass, conservation of momentum and conservation of energy are referred to as the NavierStokes equations. The nondimensional NavierStokes equations given in the previous section in the Cartesian coordinate system are now transformed to the computational space by the following. Due to similarity of the left and righthand sides of Equation (1149), the mathematical details are carried out for the lefthand side (LHS) of Equation (1149) only. Equation (1149), which is repeated here for convenience, is:
BQ BE BF BC BE" 8Fu 8C" +++=++
at 8x 8y 8z ox 8y oz
The Cartesian derivatives are replaced by Equations (115) through (118) to yield:
LHS
8Q 8Q 8Q 8Q BE BE
 8T + ~t 8~ + 7Jt 811 + (t 8( + & 8~ + 11% 81] +
{1157}
It is recognized that this equation is no longer in a conservative form. To recast the equation in a conservative form, some manipulation must be performed. To do so, Equation (1157) is first divided by J, and then a combination of terms which sums up to zero is added. In the following, only four terms are considered to show the required mathematical steps. The conclusion is then extended to the remaining terms. The first four terms of Equation (1157) divided by J and with the added zeros shown as the brackets are
Transformation of the Equations of Fluid Motion
29
1 8Q 1 8Q 1 8Q 1 8Q [ 8 1 8 1]
J 8T + J ~t 8~ + J 1Jt 81J + J (t 8( + Q 8T (J)  Q aT ( J)
which may be rearranged. as
+ [1Jt 8Q + Q!!"_ (1Jt)] + [(t 8Q + Q i_ «(t)]
J 817 87] J J 8( 8( J
[ 8 1 8 ~t a fit a (t]
 Q 8T (J) + 8~ (J) + 817 (J) + 8( (J)
(1158)
The terms in the first four brackets may be combined, and by substitution of expressions (1131) through (1134) into the last bracket, it can be shown that it is zero. Therefore, expression (1158) is combined as
Extending this conclusion to the remaining terms of Equation {1157) results in the following expression:
LHS =
8Q 8 Q 8 Q 8 Q
8T ( J) + 8~ (~t J) + 817 (T}t J) + 8( «(t J) +
8 E 8 E 8 E 8 F 8 F
ae (~3: J) + 87] (T}:z: J) + 8( «(:z: J) + 8~ (~y J) + 87] (7]!I J) +
a F 8 G 8 G 8 G
8( «(lI J) + 8e (ez J) + 871 (7]1: J) + 8( «(z J) (1159)
The RHS of Equation (1149) is reformulated in a similar fashion, and therefore
30
Chapter 11
(1149) is written as
e Q 0 [1 ]
Or (J) + O€ J (€tQ + €xE + €yF + f,G) +
:fI [~ (7]tQ + fixE + 11yF + 7]zG)] + :( [~ «tQ + (xE + (!IF + (zO)]  :€ [~ (€xE" + €yFlI + €zG,,)] + :7] [~ (flxElI + f/yF" + 7]zG,,)] +
:( [~ «». + (yFv + (zGv)]
The terms in this equation are redefined such that it may be written as
8Q aE 8E' oc aE" 8F" oo, (1160)
8T + B€ + 87] + 8( = 8€ + 8f1 + 8(
where
 Q (1161)
Q=
J
 1 (1162)
E = J (€,Q+&E+€yF+€zG)
 1 (1163)
F = J (7],Q + 7]xE + '1yF + '1zG)
 1 (1164)
G = J «tQ + CzE + (yF + (zG)
 1 (1165)
E" = J {€xEv + €yF" + €zGu}
 1 (1166)
Fu = J {fixE" + '1yFv + f1zGv}
 1 {1167}
G; = J «xEv + (yFv + (zG,,) The viscous shear stresses given by Equations (1137), (1138), (1139), (11 41), (1142) and (1144) with Stoke's hypothesis {,\ = iJ.l.} in the transformed
Transformation of the Equations of Fluid Motion
31
computational space are:
J.l [4 2
Tzz = Reoo 3(&U{ + T7z~ + (zud  3(~YV{ + T7!1v" + (yv()
 ~(eZW{ + T7zwfj + (zwd]
J.l [4 2
Tzz = Reoo 3(ezwe + T7zWfj + (zwd  3(~%ue + T7xu" + (zud
 ~({!lVe + 1]yv'I + (!lvd]
where the heat conduction terms in the computational space are:
11.4.1 Linearization
(1168)
(1169)
(1170)
(1171)
(1172)
(1173)
(1174)
(1175)
(1176)
In order to numerically solve Equation (1160), a linearization procedure is introduced, and all flux vectors are expressed in terms of the flux vector Q. The procedure was previously described in Chapter 8; however, due to its importance, it is reemphasized in this section.
32
Chapter 11
Consider the following Taylor series expansion:
 1  8E 2
E"+ = E" + 8T 6.T + G(6.T)
(1177)
In order to rewrite 8E/lh in terms of the gradient of flux vector Q, recall that E = f{Q,€t!€x'€lI'€z)
The chainrule of differentiation yields the following relation:
8E = ~ 8tj + 8E 8et + 8E 8e:z: + 8E 8ey + ,aE aez 8r 8Q 8r 8et ar o~ ar 8e" aT aez or
(1178)
For many applications, the grid is independent of time, and therefore time gradients of the metrics are zero. Hence, for a time independent grid system, Equation (l1w78) is reduced to
aE aE ao 8r = 8Q 8T
(1179)
This equation is substituted into Equation (1177) to yield:  +1  oE 80 :2 E" = E" +  .6.r + G{ar)' 8Q Br
(1180)
The partial derivative aQ / Br is approximated by a firstorder backward difference
expression as
8Q on+!  Q" 6.Q
aT = aT + D{Ar) = AT + D(Ar)
Substitution of this equation into Equation (1180) yields:
E;n+1 = E;n + ~~ [~~ + O(l!.r)] l!.r + O(l!.r)'
or
 1  aE  2
E"+ = E" +  6.Q + D{Ar) 8Q
(1181)
If a time dependent grid system is used, the additiona.l terms which appeared in Equation (1178) would be included in Equation (1181) as well. In either case, terms such as aE / oQ and similar terms related to flux vectors F, G, Ev, Fv, and Gv will always appear in the linearization process. They are defined as the Jacobian matrices, which are the focus of the next section.
For steadystate problems such as Parabolized NavierStokes equations, the linearization process is similar. In that case, the Taylor series is expanded with respect to the streamwise space coordinate (marching direction). A detailed mathematical approach is shown in Chapter 13.
Transformation of the Equations of Fluid Motion
33
11.4.2 Inviscid and Viscous Jacobian Matrices
To linearize Equation (1160), the following and similar approximations are employed:
En+l = F;n + 8E D.Q + O(D.r)2
8Q
where 8E/8Q is defined as the flux Jacobian matrix. The remaining flux Jacobian matrices are 8F /8Q and 80/8Q. The viscous Jacobian matrices are 8BIJ/8Q, 8FIJ/8Q, and 8G,J8Q. Since flux vectors Q and E are 5 x 1 vectors, each one of the Jacobian matrices are 5 x 5 (for threedimensional problems). The derivations of the Jacobians are considered next. In the following derivations we will assume perfect gas; therefore, the equation of state is expressed as
p = pe(y  1).
The inviscid Jacobian 8E/8Q is
8E 8(El, E21 Ea, B", Es)
  8(Q1, Q2, Qa, Q41 Qs)
8Q
8EI 8EI 8EI 8El 8El
8Ql  8Q3 8Q4 8Qs
8Q2
8E2 8E2 8E2 8E2 8E2
 8Q2 8Qa 8Q4 8Qs
8Ql
8Ea 8E3 8Ea 8E3 8E3 (1182)
8Ql    
8Q2 8Q3 8Q" 8Qs
8E4 8E, 8E4 8E, 8E,
=  .: = 8Qs
8Ql 8Q2 8Qa 8Q4
8Es 8Es 8Es 8Es 8Es
8Ql 8Q2 8Q3  8Qs
8Q, In order to determine the elements of matrix (1182), the flux vectors E, F, and G are expressed in terms of the components of vector Q according to
34
Chapter 11
Q2
pu Q' [lQ' Q' Q']
_2 + (1'  1) Qs __ (_2 + _1. + __!)
fYU2 + P a. 2 a. a. o.
Q2Q3
E= {lUv  Ql (1l83a)
puw Q2Q4
Q1
(pet + p)u ['YQ,  'Y  1 (Ql + Qf + ~n)] Q,
2 a. a. a, a. Q3
pv Q2Q3
Ql
puv Q' [1 Q' Q' Q']
F= pv2+p _2 + (1'  1) Qs __ (_2 + _1. + __!) (1l83b)
 o. 2 o. o. a.
pvw Q3Q4
Ql
(pet + p)v ['YQ.  'Y  1 (Q~ + til + 91)] Q3
2 a, a. o, a. Q4
pw Q2Q4
Ql
pwu Q3Q4
G= pwv  Ql (1183c)
pw2+p Q' [1 Q' Q' Q']
__! + (y _ 1) Qs  _(_2 I _1. + ___!)
Ql 2 Ql Ql Ql
(pet + p)w ['YQ'  'Y  1 (Q~ + Qf + 9!)] Q,
2 a. a. Ql Ql Transformation of the Equations of Fluid Motion
35
Recall that, from Equation (1162),
(1184)
With the components EI, Pl, and Gl from Equations (1l83a), (1l83b), and (11 83c) , (1184) may be written as
 1
El = J [€tQl + €xQ2 + €"Q3 + €zQ4J
(1185)
Now the first element of (1182) is determined as
8El 8{ ~ [etQ} + €xQ2 + €"Q3 + €zQ41}
8Ql = 8{~}
= a(etQ} + exQ2 + €"Q3 + ezQ4) = €t 8Q}
(1186)
The remaining elements of the first row are
8El
and  = O. 8Qs
The elements of the second, third, fourth, and fifth rows are determined in a similar fashion. The resulting inviscid Jacobians are:
36
Chapter 11
.....
..... I
00 ..t

+ I + + I
~s ~<:! ~.£.. I
,., ~ ,., ~ .... . ~ ,
I"',J..;' ~ 1"'1 ...... ~ t.~ ~ I
~ + ~+ ~ + ,
I ~ I ~ ~ c I
~! ~! ? ~ I ~
+ e +.., """ +.., """ I
S s S I
<:!.., <:! ..  <:..,"' 1
+ + + 1
1
1 S.. I S.., S.., I
I"":::::::"" I....:::::::.... ... ~ I
r~i~·
+ I.~ I ~I ~ I +~ ,I
'C" ~'" ~ 1 S 1 e 1  + ,
.; + < 1 I 1 I 1 ~ ~ 1
& I ~ 1 ""'1 ~ 1 ~. 1
~ 0 e.., 1"':.... I ~ I I I t 7' 1
~+~ I ""1 .... 1 e~ 110
~ S ... 1 _.. XI +''
+ '' I ~ 1 '<! ~ 1
'__' .... I ~ 1 ~ I.,... 1
"" r 1 1 I 1; 1
~~~i~+
+I""'~I ~1+'f:!>1 ~I
~~+~ .. ~II s 1 ~+ I ~ 1
& ~ I I ~~ I .', I
e ~ e.. I 1 ~ + (:3' I ~ 1
;;: + I"'..J.... 1 I 1 ~ I I I 1 ~
"S .. ~ 1 ...... 1 e ~ 1 ~ I
~..::=:.J 1 ~I +e I ~ 1
+ .... I 'e 1 ,..... <: 1
~ YI I'E; I I
:1+:::1,t
+1?~I+'f:!>1 """I ~I
'(;'"'  "'~ 1 . + e 1 ~ I
~ +1> 1 0'(;'"'1 I I I I
.s, I e I ~~I :? I :? 1
S .... +'" I + I 1 I I I 1
'" 1"'1 .... I ~ , 1 1 I ""
o <>", :? 1 e ... 1 .... I ...... 1
~~II+¥I ~I ~I
+ ,_. I"'" 1 SIS I
~ Y 1 S I 1 1
~~~~~~
~ o I ~I ::rI :? I
~ + I I I I I I I
~ I .... 1 .... I ,_. 1
~ I ~ I ~ I ~! 0
e I I 1 I
+ I I I I
~ 1 1 1 I
& I ,I
+
Transformation of the Equations of Fluid Motion
37
1 I I I ~
I I I I ~
I I I I +
I I 1 I ~'"
I... I '" I.. I ~
I;;:_ IA IA I +
1'< ...... I ....... I os
1 I 1 I r I ....
II r Ir I r I to.£:
 I ~ I  I ~ ~
·~~rr·
! I ~ I ~ ~
} I ~ I s ~ I ,.. +
_, I,.._ I ;::: ~ I ~ "'::, <t
.< I ...... , I__.....£:
I I 1 I 1 ~ I ...... I~ + .....
~ ~ I~ + 11 .... _ .... ;3
'"' 1 f't;' os 1.0'" s
1 II I ~~ I ~+ ~ ..
os I ~ I + '" I...___..... '" ~
~ I~ r ~+ I ~~ 1+
,t++
I ~', I ~
I .. I. ~
0:> e I 0:> I ~
I ~ ,.:_+ I~ 'I~~
I ;:; ~ 0:>;:, I.... ,.!; s ~
~ I I 1 ~ I 1 I ...... I~ + ::::::
I r 1;'1+ I r I' "' ...... 0:>
I  ~ ~ '"' '0:> 1 ~
I I ~~ 'I I 1.0+ r~
1;:5 + S I~ .... ~
I ~ ~+ I ~ I ~~ 1+
rtr~1
I ~ I I 1..... ~
" +~';:5 l;s 1 ...... 1 ::=:' +
~ I~ I~ I .... ;s
...!:! I ~;J I ~ I...... I ~ s ~
.... I I ~ I I I .... ,1;'1 + 
I e+ I~ I r I I .... .,~;::!
I ~ 't I I I 'I l.u + """§'
I + 8 I;J 1;3 I ~ "'os ~ ~
I ~+ I~ I ~ I ~'" 1+
1~t~+I~+I~
I a: I a: I a:
I ;3 I S I ;3 I
I + I + I + I
I "'0:> I "'0:> I ~ "'0:> I f.
I ~ + I S + I .., + I ..L
I ~ I ~ I ~ I
I + ~ I + ~ I + <'<;s I S .... ;J
~ I;J '0:>'" I 0:> X I ~+ I ~.... I.~...... I ~...... I + ~
I + ,_ I + ,_ I + ,_ I 0:> .
I ~ I ~ I ;::! ~ I ~ ......
I ;s .... 1C'l ;s .... 1C'l I ....... 1C'l I I
I ~..___. I ~..___. I ~.........._, I +
I ';( ~ I ';5' ~ I 13 ~ I ;s ~
I I + II + I + l..t+
I I I
o
38
Chapter 11
I + I + +
I ~ ~ ~ ~ 0 ~
" ~'0 ~0 ~~'€'
of'> .... s: I=! ....
~~ I ~ + ~ + ~+
"2+ I I"" I ~ lor,
w010~ 0c ~~
+ E ' ~ + .... + .... +
"\., .._.. I s:!... """" I=!.., """".. =.. """" ..
~'+" + E +
:J I ~...!, ~...  ~.. .!,
I + + +
1 E E E
1 w .. w
I ~ ._.. t '" ....
~~~0T~r~I~~l·
~ ~ ="..:r' lEI ~ 1 r:; + 1
.s, I + ~ 1 I 1 I I '::! ;:: I
=~~ .. I I ~II ~I +t..:> 1
+ t..:>1 ..... I I 1 I 1 10 I 1 10
'0 E.. '4 1 .... 1 ~ 1 ~ ~ 1
~~I 1 01 ~I +'ir' I
.... 1 = 1 = 1 0
10 y 1 liE 1
e II.._.. 1
+.LtL
+ 1  ~ I 0 I + ~ 1 0 I
~~: .. ~! ~ 10;'1 7 I
~ .!.. ~.. r 1 ~ I ~ t:; I ~ 1
.._.. + t..:>1 ..... 1 I I r 1 r 1 0
0E~1 ..... I~~I .... 1
=~II 01~~1 01
+ .... 1 e 1.,.. 1 ~ I
~ y 1 I 1; 1 ,
...:lI::~.+
+ I ? 0 1 + ~ I 0 1 0,
""~ .......... 1....+, e 1 ~ I
< +:ib 10 1 I I II
.§..I ~ I I I=!:£:, I .... 1
E ~ +w 1 + t..:> 1 ~ 1 ~ 1
t..:>1 ..... 1 ~ I, I 1 I 1 0
0'E ~ 1 e ~ 1 .... 1 ..... 1
S:~II+~I 0101
+ ..... I "" I S lEI
~ '' I.!, I I I
J..T';
~ ot:'..l I 1 .... I
0+ I ~ 1 ~ I ~ I
s:! 1 ..... I ..... 1 ..... I
~ I ~ I 0 I 0 I 0
e I I I 1
+, I I I
~ I 1 1 1
.E. I 1 I I
 I I I I
+
Transformation of the Equations of Fluid Motion
39
Some of the terms appearing in the Jacobian matrices may be defined such that they are expressed in a compact form. For example,
(1l90a) (1l90b) (1l9Oc)
are known as the contravariant velocity components. They represent velocity components which are perpendicular to planes of constant e, T/, and (, respectively. FUrthermore, define
(1191)
Before attempting to determine the viscous J acobians, the viscous flux vectors must be rearranged by substituting expressions for shear stress and heat conduction terms. For example, consider the second component of E1J, which from Equation (1165) is given by
 1
E112 = J (~XE112 + eyFlI2 + ~zGlI2)
From Equations (1152), (1154) and (1156), the following is observed:
(1192)
Hence, Equation (1192) may be expressed as
(1193)
Now, using the expressions for the shear stresses given by Equations (1168), (1171) and (1172), Equation (1193) is rearranged as
+ (yVd  ~J1.(eZW{ + T/.zw'I + (zwd] + ey [J1.(eyUe + 1]y~ + + (yU, + &ve + T/xV,., + (xV.:)] + €z [JL{€zUe + T/z~ +
+ (zu, + exwe + T/xWfI + (xw.:)] }
40
Chapter 11
1 4 2
(3~z€z)we + (3€zl1z + €IITiIl + €zTiZ)1lr] + (€II7JZ  3&7JII)V'1 +
2 4
(~zl1z  3&7Jz)w" + (3€z(z + ~II(II + €z(z)u( +
All the components of flux vector El1 are reformulated accordingly. In addition, viscous flux vectors Fv and 011 are modified in a similar fa.shion, resulting in the following form of the viscous flux vectors:
Transformation of the Equations of Fluid Motion
41
o
where, for a perfect gas,
1 'Y [ 1( 2 2 2]
Pr{"t  l)M! T = Pr et  2 u + v + w )
(1194)
(1195)
42
Chapter 11
o
(1196)
Transformation of the Equations of Fluid Motion
43
o
.
+ ~h(v2)'1 + ~!a(W2)'1 + fsuv'l + fsuw'l + frv~ + fsvw'l
(1197)
44 Chapter 11
where
a = 1e+e+F,_2 (1198) a =e+ 1F,_2+e (1199)
1 3 z ~ % 2 % 3 ~ %
es = €; + €; + ~€: (11100) a =€2 +e+t2 (11101)
4 % ~ .• %
as = l€z€t (11102) tl6 = i€v€% (11103)
a7 = i€z€1I (11104)
b = 1T]2 + "12 + "12 (11105) b2 = ,tz + ~11~ + 11~ (11106)
1 3% Y II
ba = r& + T/: + ~11: (11107) b4 = f1~ + 11: + '11: (11108)
bs = ll1z'hr (11109) be = l11117J% (11110)
br = il1%7J% (11111)
C = 1(2 + (.2 + (2 (11112) C = (2 + 1(.2 + (2 (11113)
1 3 % ~ z 2 % 3 11 z
_ (2 (.2 + 4(2 (11114) C4 = (~ + (; + (; (11115)
C3%+y 3Z
Cs = l(z(v (11116) C6 = l(1f(z (11117)
C7 = l(%(% (11118)
d1 = t€z11z + €1f1'11l + €%7J% (11119) ~ = €%T}r + ~€y1/1I + €%T/z (11120)
d3 = €z7J:J: + €1fT}1f + ~€zT}z (11121) d4 = €rf"J% + €yTI" + €zT/z (11122)
ds = €%77y  i€y7Jz (11123) ck = €r7Jz  j€x1/z (11124)
d7 = €y7Jz  i€r1'11l (11125) d8 = €1f1]z  1€1l1'11l (11126)
dg = €zT/%  i€zT/z (11127) dlO = €%1]y  i(y1/% (11128) Transformation of the Equations of Fluid Motion
45
el = ~~xCr + ~y(y + ~;{z (11129) e2 = ~x(x + ~~1I(1I + ~z(z (11130)
e3 = ~r(x + ~Y(lI + ~~z(z (11131) e4 = ~%(x + ~1I(1I + ~z(z (11132)
es = ~x(lI  i~lI(x (11133) ~ = ex(z  iez(r (11134)
er = ~yCx  ~~x(y (11135) es = ey(z  i~z(y (11136)
eg = ~z(x  ~ex(z (11137) elO = ~z(y  ~~11(% (11138)
11 = ~17r(x + 17y(y + 1721(% (11139) h = 17x(% + ~17Y(Y + 1721(% (11140)
fa = 17r(,: + 17y(y + ~17z(% (11141) 14 = 1]x(x + TJy(y + 17%(% (11142)
/s = 17x(y  ~17y(x (11143) /6 = 17r(z  ~17%(x (11144)
h = 17yCx i17x(y (11145) fs = 1]y(%  ~ 17%(y (11146)
fg = 17%(.  ~17x(z (11147) flO = 17%(11  ~ 1711(% (11148) The derivation of viscous Jacobian matrices must be handled with special care, due to the fact that the components of the viscous flux vectors involve gradients of the dependent variable and the Jacobian of transformation, which is itself a function of the independent variables. Thus, J remains embedded inside the viscous Jacobians, unlike the inviscid Jacobians where the Js were cancelled. In order to generalize the mathematical procedure, it is recognized that all of the terms in the viscous flux vector may be expressed as
 1
Et = (Factor) J(¢J",) ,
(11149)
where ¢ represents the dependent variable such as u, or a combination of dependent variables such as u2 and 'If; represent an independent variable such as ~, 17, or (. For example, the first term of the second component of s; (from Equation (1194)) is
 IJ. 1
Ev2,1st term = Reoo Ja1(ue)
which may be expressed as
46
Chapter 11
or, in the general formulation of (11149),
 1
Eu2,lT = (Factor) JtP¢
To obtain the viscous Jacobian 8Eu/8Q, it is necessary to compute each of the terms in 8(Eub EIJ2, Eu31 EtJ41 EfJs)/8(Qll Q2, Q31 Q4J.JS)' In a similar fashion, the Ja.cobians associated with flux vectors Fu and Gu may be determined. In order to perform this differentiation, the general expression (11149) will be used to illustrate the procedure. Consider the determination of
1 8 84>  (Factor) ( J) 8Q 8'1j;
At this point the order of differentiation is interchanged to yield:
(11150)
Recall that Q = Q/ J. Hence, after substituting into (11150), one obtains
(11151)
This formulation is used to obtain the viscous Jacobians given by Equations (11152), (11153), and (11154):
0 0 0 0 0
EQ2,1 EQ2,2 EQ2,3 EQ2,4 0
8Ev JL EQ3,1 EQ3,2 EQ3,3 EQ3,4 0 (11152)
~=
8Q ReooJ
EQ4,1 EQ4,2 EQ4,3 EQ4,4 0
EQS,I EQS,2 EQS,3 EQS,4 BQs,s Transformation of the Equations of Fluid Motion
47
0 0 0 0 0
FQ2,1 FQ2,2 FQ2;3 FQ2,4 0
8ft!) J.L FQ3,l FQ3,2 FQa,3 FQ34 0 (11153)
... 
8Q ReooJ ,
FQ4,l FQ4,2 FQ4,3 FQ4,4 0
FQS,1 FQS,2 FQs,3 FQS,4 FQs,s 0 0 0 0 0
GQ2,1 GQ2,2 GQ2,3 GQ2,4 0
8C!) It GQ3,1 GQ3,2 GQ3,3 GQ3,4 0 (11154)
8Q = ReooJ
GQ4,l GQ4,2 GQ4,3 GQ4,4 0
GQS,l GQS,2 GQS,3 GQS,4 CQs,s
where
[u v w 1
EQ2,1   a}(J)e + as(J)e + a1(J)e
p p p
[u v w 1
 d1(J)fl + dr(J)fl + dg(J)fl
P P P
[u v w 1
 e}(Jk+er(J)(+eg(Jk
p p p 48
Chapter 11
[ u v w 1
 es(Jk+e2(Jk+e1O(J)c
p p p
J J J
EQa,2  as()e + ds()'1 + es(k
p p P
J J J
EQ3,3  ~()E + d2()'1 + e2()'
p p P
J J J
EQ3,4  Cl6( )e + dlO( )'1 + e1O( ),
p p p [u v w 1
a1(J )e + l16(J )e + a3(J )e
p p p
[u v w 1
 rk(J)'1 + ds(J)'1 + d3(J)'1
p P P
[u v w 1
 es(Jk + es(Jk + ea(J)c
p p p
Transformation of the Eq uations of Fluid Motion
49
J J J
 ~()e + ds( )'1 + es(k
p p p
EQS,l
v w J
+ (d7u + dsw)(J)" + (~u + dlOV)(J)f} + (d5v + c4;w)uf}
P P P
J J } { l'e4 u2
+ (d7u + dsw)v'I + (dgu + dlOV)W'I  (e,  p )(J)c;
p p r p
u v w
+ (esv + e6W) (J)c; + (e7u + esw)(Jk + (egu + elOV) (Jk
p p p
J J J }
+ (esv + ~w)u, + (e7u + esw)v, + (egU + elOv)w,
p p p
50
Chapter 11
EQs,2
EQS,3
EQS,4
EQs,s
[ 'Y u J J J 1
+ (d1  p d..)(J)f) + (dsv + t4w)()f) + d7()vf) + dg()Wf)
r p p p p
[ '"Y u J J J 1
+ (el  p e4)(J ),; + (eSv + t%W)( )c + e7(  )v( + eg(  )W(
r p p p p
[ '"Y v J J J 1
+ (d2  p d..)(J)f) + (d7u + dlOW)()'1 + ds()~ + dlO()W'1
r p p p p
[ f v J J J 1
+ (e2  p e4)(J)" + (e7U + esw) (), + eS(')U( + etO( )W(
r p p p p
[ '"Y w J J J 1
+ (d3  p d..)(J )" + (dgu + dtov)( )'1 + lis ( )~ + ds(  )v"
r p p p p
.. 
Transformation of the Equations of Fluid Motion
51
J J J
FQ2,2  d, (  h· + b1 ( )'1 + 11 (  Ic
p p P
FQ2,3 J J. J
 d5( )€ + b5( )'1 + h( ),
p p P
FQ2,4 J J J
 ds( )e + b,( )" + jg( ),
p p p J J J
 d7()e+b5()'1+/5(),
p p p
J J J
 ds( )e + bti( )" + llO( ),
p p p
.
ODTO KOTi)PHANESl M. E. T. U. LIBRARY
52
Chapter 11
FQS,l
[ U v w 1
 br(J)" + b6(J)" + b:J(J)"
p p p
[u v w 1
 f6(J), + fs(Jk+ h(J),
p p p
J J J
 dlO()e+b6(),,+!s(k
p p p
I V2 I W2 I e
+ (~ b4)(J) + (b:J  b4)(J) + b4(J~)
Pr p " Pr p" Pr p "
uv VW UW } { I U2
+ 2bs(J)" + 2bs(J)" + 2b7(J)"  (/1  p f4)(J), +
p p p r p
Transformation of the Equations of Fluid Motion
53
FQS,2
FQS,3
FQs,4
[ 1 v U w 1
+ (b2p b4)(J)'l+bs(J)'l+b6(J)'l
r p p p
[ 1 v J J J 1
+ (12  p f4)(J ), + (!1u + fsw)(  k + Is(  )u( + ito (  )w(
r p p p p
J 1 [ 1 w u vl
+ dlO()V~ + (b:ip b.)(J)'1+br(J)'1+bs(J)'1
p r p p p
[ 1 w J J J]
+ (fs p f4)(J)(+ (fgU+ftoV)(),+/6()U,+ fs()v,
r p p p p
54
Chapter 11
FQs,s
[u v w 1
GQ2,l   el(J )e + e5(J)e + Cf;(J)e
p p p
[u v w 1
 h(J)" + fs(J)'1 + f6(J)"
p p p
[u v w 1
 Cl(J),;+es(J),+C7(J),;
p p P
J J J
GQ2,2  el()e + ft()'1 + Cl(),
p p P
J J J
GQ2,3  es( )e + fs( )'1 + C5(  k
p p P
J J J
GQ2,4  e6( )e + f6( )'1 + C7(  k
p p p [u v w 1
 h(J )'1 + h(J )'1 + Is(J )'1
P P P
[u v w 1
 es(J),; + c2(Jk + Ct3(J k
p p p
Transformation of the Equations of Fluid Motion
55
[ u v w 1
eg(J)e + eIO(J)e· + e3(J)e
p p p
J J J
GQ4,2  e9()e + 19()" + C7(),
p p P
GQ4,3 J J J
 elO( )e + floe )" + (1)( ),
p p P
J J J
GQ4,4  e3( )e + h()" + C3()'
p p p GQS,l
, .
56
Chapter 11
GQs,2
GQs,3
GQs, ..
u v W
+ (/7v + fgW)(J)'1 + (fsu + !lOW)(J)'1 + (fsu + JsV)(J)'1
P P P
J J J }
+ (/7v + hW)14J + (fsu + !lOW)V'l + (f6U + fsv)w'1
P P P
J 1 [ 'Y u J J
+ es(  )we + (It  p h)(J )" + (/Tv + fgw)( )" + Js(  )v'1
p r p p p
[ 'Y v J J J 1
+ (f2  p ! .. )(J )'/ + (fsu + JlOW) ( )" + h( )~ + !s(  )w,/
r p p p p
Transformation of the Eq uations of Fluid Motion
57
11.5 ThinLayer Approximation
The NavierStokes equations are reduced rather drastically by assuming thinlayer approximation. Under this assumption the gradients of the viscous stresses in the direction parallel to the surface (~ and ( directions) "are neglected. The ThinLayer NavierStokes equations are expressed as
B(j BE BF Be BFVT aT + B~ + B." + B( = 7frI
(11155)
where flux vectors Q, E, F, and G are given by (1161), (1162), (1163), (1164) and
lb1(U2)'1 + ~~(V2)'1 + lb:J(w2)'1 + bs(uv)'1 + b6(vw)'1 + lrr(uw}'1 + prh~l)M!, b4T'I
(11156)
o
The viscous flux vector FUT is obtained from the viscous flux vector Fv given as Equation (1196) by omitting all the mixed partial derivatives. To illustrate this
58
Chapter 11
point, consider for example the second component of the flux vector FIJI where
The partial differential equation (11155) includes the 11 gradient of the viscous terms. Therefore, this gradient will include terms such as 82u/8118e, 82v/8118e, ... and so on. These and similar mixed partial derivatives are omitted. As a result, only 11 derivatives of the flow variables remain, which ha.ve been redefined as the flux vector PVT' The reason these terms are dropped will be discussed in Chapter 14 where the numerical solution techniques for thinlayer and NavierStokes equations are considered. The elimination of these terms is not necessary; however, this reduction does simplify the equations and reduces computer storage and time requirements.
The inviscid Jacobian matrices are the same as those of the NavierStokes equations, and the Jacobian matrix 8FvT/8Q is
·.~ .
. "~.
Transformation of the Equations of FLuid Motion
59
I "" "" ~ ;:II
I ,., .... ..........
""(<::1. ""(<::1. ""(<::1.
'"  
0 ~ .s ~
;:II
.,. .,. ~ rl
.......... .... .... I
""1<::1. '"')1<::1. ""1<::1.
._ '' ..__.
0 .$ .s .s
 
~.~ ..
~
~I<::I.I ~I<::I. .....
s I <::I.
..., '"') ...,
._ ._ ''
..li I .s .e
+ + +
~ "" s
.... .... ;;I<=l.. I "';:II
;:.1 <=l.. ;:.1 <=l..
..., '"') ...,
''"  .__.
~ ..s .2
+ + + rl
I ~ I .,. ..f:
;:11<::1.1 
I ;:I I <=l.. ;:II <=l.. I
I i I_.....!::
I I I """:.1 <::I.
I , I '"
I I ~
= ] 0 I 0 I 0 I rlct
l',tr _.....!::
I I I I ~~<=l..
I I I I ':i
I I I I ..f: rid:
I ~ I .,. I, .,. I .s; ;:.1<::1. I
I .......... I.......... .......... I ;:II <=l.. ...,
I '"':.!..<=l..1 ~<::I.l~<=l..I::, ~ e
01 ..li1.4IJfI...o + +
·rrt+,
I I I A
I I I ;:.1 <=l..
I I I ...,
I I I ;:::::
I I I "
I I I ..Q
II I I rltt ..f:
, ..f: 1 ~I<=l..
<::I. "":.
'"') ..8 ._, .__. '' .J§
~ + +
_._
,
.,.
..,
.,. .,. ~ I <=l..
"';:.Ic:..."'~Tc:... ::,
.2
~ """:. C'l
._ .._, +
.._ ... .,.
~ ~ .... .,. ..f:
rl~ rl~ .rl<=l..  S I <=l..
..., EI<=l..
.._. ;:I,
I I ~ '"') '"')
.__. .._.
'" Jf <Id: ~ .£
~ ._ C'l C'l
+ + + + + ~Id
II
1~llg
60 Chapter 11
11.6 Parabolized NavierStokes Equations
For steady supersonic flow fields, PNS equations provide an efficient method of solution. It is assumed that the streamwise gradient of viscous stress is small compared to the gradients of stresses in the 1] and (directions. Therefore, the streamwise gradient of viscous stress is dropped. In addition, the streamwise pressure gradient within the subsonic portion of the boundary layer must be approximated to prevent a departure solution. This approximation may employ anyone of the following methods:
a. Neglect the pressure gradient completely.
b. Treat the pressure gradient explicitly.
c. Impose the pressure from the first supersonic pointthis method is known as the sublayer approximation.
d. Retain a fraction w of the pressure gradient.
The PNS approximations will exclude streamwise flow separation; however, cross flow separation is predicted. In the following formulation of the PNS equation, the streamwise pressure gradient within the subsonic portion is approximated by method d. The PNS equations are expressed as
aEp aEpp of oc aFtJP aGIJP (11157)
a~ + ar + a1] + ae = 8T1 + 8(
where
Ep 1 (11158)
 J [€~Ep + €JlFp + €zGp]
Epp 1 (11159)
 J [~~Epp + ~JlFpp + ~zGpp]
F 1 (11160)
 J [1]~E + 1JyF + 1JzG]
G 1 (11161)
 J [(~E + (yF + (zG]
Fup 1 (11162)
 J [1]:z;Eu + 1JyFtJ + 1]zGtJ]
GIJp 1 (11163)
 J [(:z;EtJ + (yFtJ + (.,Gu) Transformation of the Eq uations of Fluid Motion
61
and flux vectors are defined as
pu
pu2+wp
Ep= puv (11165)
puw
{pet + p)u pv pvu
Fp = pv2 +wp
pvw
(pet + p)v
(11166)
pw 0
pwu (1 w)p
Gp= pwv (11167) Epp= 0 (11168)
pw2+Wp 0
{pet + p}w 0 0 0
0 0
Fpp= {1 w)p (11169) Gpp= 0 (11170)
0 (1w)p
0 0 62
Chapter 11
The flux vectors E, F, and G are the same as Ep, Fp, and Gp with w = 1. The parameter w, introduced in Equations (11165) through (11170), is determined by stability analysis. The viscous flux vectors Fup and CuP are
where
o
~bl(U2)" + ~~(V2)" + ~~(W2)" + bs(uv)" + ~(uw)'1 1
+br(uw)" + Pr(,,( _ l)M~ b.T"
o
(11171)
Transformation of the Equations of Fluid Motion
63
b  4 2 2 2 (11173)
1  "317% + 17" + 17%
b  2+ 2+42 (11175)
3  17r 171/ 3"'1/
bs = ~17%171/ (11177)
b7 = ~17x17% (11179)
_ 4(2 (2 (2 (11180)
Cl  3" % + 1/ + %
C = (2 + (2 + 1(2 (11182)
3 % 1/ 3%
Cs = k(:r;(y (11184)
C7 = i(:r;(% (11186) b = 1]2 + i'J12 + 'J12
2 % 3'111 '1%
(11174)
(11176)
(11178)
(11181)
(11183)
(11185)
Note that the viscous flux vectors Fup and Gup are obtained from relations (11 96) and (1197) by omitting all the mixed partial derivatives, as was illustrated previously for the ThinLayer NavierStokes equations,
The inviscid and viscous Jacobian matrices for the PNS equations are as follows:
64
Chapter 11
. .... .... I

00
.....,J

+
+
+
+ ~
+ e
..,
+
o
Transformation of the Equations of Fluid Motion
65
,.. 00 00
.......
I ....... ....... ''
" .
66
Chapter 11
, , + I
I '0 s +
I .....'' ~ ~
I' t:l,....... €" ,...
~ ... ~0
1 ,+ ::;" ~
, ....... ~ I I + ~
17<1 ~'\;' I
~
.. if'" , ,L~''''': ..  I ~
," ,_ L_~ I
., ; ~ I ';,' c .
....... <.:!'" _ I 1 II ~ 1 ! E:::tI, 
" + ""_ 1::; 1 " '" ,
• ~_::; I ::; I' ~ I I
+ = I  I ~ ,
_ " 0' ",. 1
'" " ",I < ,"
___ ' __ I I =: I + I
.. ~ , 1 t '0 '
• • ,__  I
....' ~  ~ ,
+ >,' 
e ~ I ~ + 0 I 
• I I r;:; ., '"
+ ""  ::;'" I "
s_ ::; I ! + > I
= I  ' '" ¥ ::;
....... __ I e I
'7" ~ +
:___
'\;' > ~1
0'
=:
+ :0
~ I
+ I
'" I
___ =r;~
+
'0 .....
.§_ 4
I
~ .....
0'
i'!
+
~
<.:! + 
'0 ,.. ~
.§... 4 ..,
I +
s ....... <::
.__.. ..
0' + t:ll .....
t! ~ ..
.. 4
+ ...::::::... I . .....
..... I .....
OQ
$
+
+
o
+
e ...
+ S
...
...::::::...
''"
L~ +

~.
'0 ~
I
::?
I
..... ._,
c
o
" '
Transformation of the Equations of Fluid Motion
67
I i ~
I I I _......_.
II I I I '";) I <=l...
I I I ..__....
I I I I ~~
o I 0 I 0 I 0 I ?Io...
1~~,
~ _......_.
s I Q.. ...., ..__....
.. o 0'1

I
..... ..... .._,
" .
68
Chapter 11
Q)~ ~I
<0:;,
II
tl~
0
? ("I ~;' ("I
... ...
.... ........... ~
I <, t ... ~
:PI~ "l;) I ~ "l;) I ~ "l;) I ~
....___,., ....___... ....___,.,
,.. ..... '"
i! + + +
...._... s s s
~
~ ~ ........... ~
"l;) I~.., to.,. c ... to.,.
"l;) I~ "l;) I~ "l;) I e
....___,., ....___,., ....___... ....___,.,
,.. .... .... ,..
+ I + .... ~
I "l;) IS
l~
+ ~r~
~ ... 1...........
I I"l;) I to.,.
I ....___,.,
~I~ I '"
~ 1
...._... I
........... 1 (.., I
~ _l______ TI~
~I~ 1;;I
2 I
~II
"l;) I ~ I ....___,., I
: ~
,..... I 8 1
I
+
~
+ ,..... ("I
..,
+ Q
+ ~
+ s
+ S!
+
("I ...
_... 

I

\0


+
J~ I !l
I~I~ I
~~ I'~ t~·
,.. ...
:;~ II ~ I
....___..., ....___,.,
.... I ,.. I
! I
I I
I I
I I
I I
I I
,:;t;i,~
•• ... I
;:"j":::' ~ I
 ....___... I
,.. "" I
I
I
,
I I
I I
I I
I I
I.t_J_
8 ,I{ I ~ 10
~~SIf31 I
I
I
1+1
~;; ~ a
I \
" .
Transformation of the Equations of Fluid Motion
69
11.7 TwoDimensional Planar or Axisymmetric Formulation
For a twodimensional planar or axisymmetric flow field, the nondimensionalized NavierStokes equation may be expressed in a combined form as
8Q 8E 8F 8E" 8F"
 +  +  + aH =  +  + aH"
at ax ay ax ay
(11192)
where a = 0 represents twodimensional planar flow and a = 1 represents twodimensional axisymmetric flow. The flux vectors in Equation (11192) are
p pu
Q= pu (1l193a) E= pu2+p (1l193b)
fYV puv
pet (pet + p)u
fYV pv
puv (1l193c) 1 puv (1l193d)
F= H=
fYV2 + p Y pv2
(pet + p)v (pet + p)v
0 0
E,,= Tzxp (11193e) Fv = Tzy (11193f)
Tzy TWp
UTxxp + VTxy  qx UTxy + VT!/lIp  qy
0 2 Y a v T     ("' ) xy 3 ne; 8x y
1 2 '" v y 2 a v
H" =  TyllP  TOO  3" Reoo (li)  Reoo 3" ay ("'1)
y 
2 '" v2 y a 2 v2
ur + VT.  q        ("')
xy YW 11 3 Reoo u ne; ay 3 y
u 8 2 uv
 Reoo 8x (3"'y)
(1l193g)
70
Chapter 11
where
(11194)
TlIlIP = .E: (~ {}v _ ~ au) ne; 3 {}y 3 ax
(11195)
(11196)
J1. [ 2 Bu {}v 4 v 1
TOO =   ( + ) +  
se; 3 {}x {}y 3 y
(11197)
J1. 8T
ReooPrb  I)M;" ox
(11198)
J.L {}T
ReooPr('Y  I)M! lJy
(11199)
The NavierStokes equations given by Equation (11192) are transformed to 'the computational space by relations (116) and (117), resulting in
(11200)
where
 Q Q=J
(11201)
(11202)
(11203)
 H H=J
(11204)
(11205)
Transformation of the Equations of Fluid Motion
71
 HI)
H=I) J
(11206)
(11207)
Substitution of viscous shear stresses into expressions (11205) and (11206) provides
o
~al(U2)e + ~a2(v2)e + aa(UV)e + pr('Y~l)Ml a,Te
+~Cl (u2)'1 + ~C2( v2)'1 + cauv'I + C4Vu,., + Pr(y~l)M! csTfJ
and
o
~Cl (u2)e + ~C2( v2)e + C3VUe + C4UVe + Pr('Y~l)M!. csTe +kbl(U2)'1 + 4b2(V2)'1 + b3(uv)" + Pr('Y~l)M!' b,T'I
where
4e+e (1121Oa) a2 = ~~ + t~; (1l210b)
al  3 x y
a3 = i~x~y (1121Oc) a, =~; + ~~ (1l210d)
b = 1.fJ2 + fJ2 (1l211a) b  2 +' 2 (1l211b)
1 3 x 11 2  fJx 3"1y
bs = IfJx1Jy (11211c) b4 = fJ~ + 17; (ll211d)
Cl = ~1Jx~x + ~yfJy (11212a) C2 = {x1Jx + ~~uTN (1l212b)
Ca = 1Jx~y  j~xl7l1 (11212c) C!.i = ~zl1l1  ~~ul1z (1l212d)
Cs = ~z1Jz + ~yT}y (11212e) (11208)
(11209)
72
Chapter 11
The inviscid and viscous Jacobian matrices are:
t.:), ...... ".....
'<:
 ~
~
I
.....
..._,
~
~
IW
+
~
IW .

I
tv
......
w

..._,
~1~~
"""'1 I 10
 1 ~ I ~ 1
t: I I I I 1
< I ...... I ..... 1
~ I ~ 1 ~ 1
~ I 1 1
+ I I 1
o I 1 I
~ I I 1
 I I
Transformation of the Equations of Fluid Motion
73
74
Chapter 11
C) 11
2)~ ~I
I
+
..........
~ I I
!. c:t, g I 0
:::: I
~w I
+ I
~ I
...:: I
._. I
rLt.
I I 'I I
.......... I I
~ I I I
I I 0 I ~ I 0
.  I I
 I I I
~ I I
____ ~ __ ! + .L . _
~
C1l
...
.,......
~I! ~
<c::» ~
...
~
~
W + I
~ I
W I
2:""...1++.
I I I
I I I
I I I
I I I
I 0 I 0 I
I I I
I I I
I I I
I I I
c.
Transformation of the Equations of Fluid Motion
75
I~
I ~
I I t!
I I ?II..
~~f~~'~ 
I ~
II ~
I ~
I~~~~~ ;
I "":> I ~.....__, ..., I ~ .._____.,. I ......:::._,
I"__""..., ..__.... ... I ...,
1 .... u co \) 1:1
01 1:1 + 1:1 + L......l
~__1_~
i I I
I I I
I I I
! !
~I ~I ~
'"" ,...... '" ........ I ;:. I ~
e
~
""I~
..__....
Cf
?I~
+
.... ...
~ ;:
 ., ,......
;:>1 ~ .._, ;:>1 ~
..., ""I~ ...,
..__.... J .._____.,.
,...... + ,......
... s
1:1 ;:
?I~ ..... tI~
""I~
I ...____... I
'" ., '"
1:1 (j ~
"'"
+ + +
  ""I~ ..... ..." ~ I ..., ;:.
""I~ .....__, "":>I~ ...____... I ....__...., rr J ~
....__...., .... ..__... J ..., "'"
1 \) ..., t! + +
~; + 1:1 + +
  
'""

sl e,
...,
""""
...,
1:1
C'l ;:
.., '"" .... + ;: ;I
.... ;: ~ ;: .....
,,...., ...    .... ~ ~ ~I ~ 
., 1 ~ ., I ~ ~I e, ~I ~ "';II e, ..... "'.,1 ~ ... "":> "'I~
..., ..., ..., ..., ..., 04;:.1 ~ ..., "';:>1 ~  ...,
....__.....    ....__....,
..__... ...,  ..., ;:>
'" {J '" J ____.... ____.....__  ... J ____.....__
" 1:1 ,...... ..... u
+ + + + ... "..... .... .., + ;:
t!  t.l ..... + 
?Id: ... ?jd: s
.... ~ '"" ~ t! UI<=I. ;: ;: .sl<=l.
    ?Id: ?Id:  ;:>
;II ~ ~ 1 <=I. P I <=I. ;:>I~ ..., ., I <=I. ..... "":>
, I  I 
..., ~ ..., ~ ~ ;I I <=I.
  .... I ... ~ I ..., is
... ~ ... t! 8 ....__...., tI~
ej ... e ?Id: ;:l
~ t! t!
..... '_.;' ...____.. '"" {J {J
0 I I I I I + + I +. + + + rI ....
.....
= I <=I.
~
......
J
rid:
I
76
Chapter 11
+ + + + +
:P14 ...."0. ~14 n ......
...... ....
s: ...... ~ n
..,
s: 1 :r s "....._, 1
t.....
"....._, :P14 I "....._, ~le :P14
<:..... :P14 t..... ....___"
~I.t' 2" ~I~ ..... n
..._., ...__:.... + ...
'" s: 
.;i ,..
'v' ",...... '" ......_.,...._. & "......,
t..... "....._, "....._, <:.....
~I<::'" t..... <:..... ~Io:t.
__... ~ I~.., ~ I~ ..._.,
.;i __... __... ......
+ '" ~ +
.....
t:) + &
3" ~
",...... ~ "......,
<:..... ....... t.....
~ I§ t..... ~ I~
~ I~ __...
__... __... .....
.;i
e
..... I
1 I 1 0
...."0.1 3" rn' s: S'
71 (ol
"....._, ,.......... ....... "....._,
1 I t..... <:..... t..... c....
~ I>:: ~, I>:: ~ I>:: ~ I~
~14 I __... ...... __... ......__., ..._.,
.;i ,., .:s
~ + + + +
 s: s s: ~
..... I "....._, ..... ..... ..... "....._,
e..... I
~I;:: I c.... <:..... <..... <.....
~ Ie "l::. Ie ~Ie ~ t e
...__:; I ......__., ...... __... '' ''
,., .:::I ..... .:::I ,.,
LI .t
I
I I
~I
o
+
......
s:
1
~14
s:
'"
"....._,
t.....
~I;::
......__.,
'"
+
......
0
.,
. I
 :P14

I
tv 2"
 '"
..,J
 .......
t.....
""=:lIe
....___.....
.:::I + +
s: ......
~
..... 1
<:.....
"I::> t e ::.014
......__.,
.:::I ~
 + ~ + S'
s: 0
.... ... ... .......
"......, "1:) 1<.....  "1::>11'......
~I<..... "_" ~I<..... ..._.,
......__., ..... ......_... ....
.:::I .:::I +
I I
I I
t·f+
+ + o I + :J I + ~ 0
...... (ol y .... ...... I {f ....
s: :J ~1t.....1"""'_' "1:)1<:.....1.... ~It.....
.... I ~ I ~ It..... ,_....,. I "I::> It..... ......_...
I'...... I......__... I ......__... ....
~ ::.014 t I _" I .:::I
.:::I s ;:: I I
 .... I I
~~I I
·"tl~I_,  4'1:tl·~110 To
"'I s I
r "......, I
~~I I I
" .
Transformation of the Equations of Fluid Motion
77
0 0 0 0
HQ2,1 HQ2,2 HQ2,3 0
1
GI}= (11218)
ReooJy
HQa,l HQ3,2 HQ3,3 0
HQ4,1 HQ42 HQ4,3 HQ44
I ,
where J 2 [J J.t] J 2 [J J.t]
HQ2,3 = ~x( )e  3Y(.z ()( ) + J.tTJz( )"  3YTJz ()()
P PYe P PYl1
[V] 2 [J.t V] [ v J V]
HQ31 =211 (.!/J)e +3Y(.Y ()(J) +211 rh .. (J)l1+()
, P Y pe P YP
2 [J.L V]
+ Y17 ()(J)
3 Y Y p"
HQa,2 = 0
78
Chapter 11
"'t I' [ et ( J 2 J 2)] (U ) [ ]
+ P e,  J + 1.£ + v I' J T1z;V" + TIsr~
r P P PeP
2 [ v U] 4 J.L v2 4 [J.L V2]
+J.Lv 2n(J) +T1z;(J) +(J)+YTl ()(J)
3 "IJI o" p" 3 Y P 3' Y p"
4 [ J.L UV] "'ttt [ et J 2 J 2 ]
+ YT1:e ()(J) + 1]y  J + (u + v )
3 'II p" Pr P p P "
J J 2 J "'tJ.L u
+ J.L()(fkV" + 17yu,,) + ttu1]y()"  3J.LvTl:e()"  P ey(J)e
p P P r P
2 [ J.L V] Itt u
 YTJz; ()(J)  rt (J)
3 Y os« Pr 1/ o"
4 [J.L V] 2 [tt U] Itt v J
3yell ()(J) 3ye:r ()(J) P ey(J)e+J.LUrtz;()"
P Pe 'II Pe r P P
2J 4 J 4J.L.v 4 [J.L V]
+J.L(2nv 17:e7l·)+ILV'Yl.() (J)YTl ()(J) +
3 P "IJI" '1 3 ·IJI o" 3 y P 3' Y as«
Transformation of the Equations of Fluid Motion
79
2 [J.i U] 'J.i v
 Y11x ()(J)  Y (J)
3 Y P TI Pr TI p '1
The ThinLayer NavierStokes equation for twodimensional planar or axisymmetric flow is reduced to
8Q 8E of  8FlIT 
8T + 8~ + 811 +aH=~+aHlI
where flux vectors Q, E, and F are given by (11201), (11202) and (11203), and
(11219)
 J.i 
FliT = ReooJ
b3U'l + ~v'l
(11220)
o
The Jacobian matrices aE/aQ and 8F/at;} are given by (11213) and (11214), and
80
Chapter 11
..


I
N N
 '"
~~,
~
II
tl~
+ + + 0
t.:> ~I~ ",.... 
0 :r s: 0
(0) ",.... ....
_,......." ~ I 0 _,......." _,......."
...
~  I ~ ~
~ Ig ~ ~I~ "l::ll~ "l::ll ~
~I.t' :PI~ '"" '""
...._.... ~ .:s .:s
.:s ....___... ......__" s: + +
...__." .:s
......... ''" S'" 8'
t...., .....
~ I~N ~ _,......." ,.......
~ I~N ~ ~
...._.... ~I~ ~I~
.:s ....___... .......___.,., ....___...
.:s .:s .:s .t+
+ ",.... s: s: I 0
(%" .9" ....... ....... I
I ~I~ '1:I1~ I
I '"".......___.,., I
~~..:s .:s I
? I
''" I
....... I
~ I
~I~ I
'"" I
.:s I
i ... ~~
+ ~ l:r s: 0
",.... ..... 1""'" .......
S'" t...., I "t:ll~ ~It....,
I ~I ~ ;
:P1~.:s I
? I
''" I
~ I
~Ic::! I
'"" I
. .: ~l._I
I I I
::PI~ I 0 I 0 I 0
• I I I
s: I I I
....... I I I
"I:l I t...., I I I
......___.... I I I
.:s I I
I I
Transformation of the Eq uations of Fluid M otion
81
The Parabolized N avierStokes equations under the assumptions stated previously are formulated as
8Ep oEpp of fI _ BF"p fi
Be + Be + 01J + Q  01J + a II
(11222)
where
(11223)
(11224)
and
pu
pu2 +wp Ep=
puv
(pet + p)u
(1l225a)
pv puv
Fp = (1l225b)
pv2 +wp
(pet + p)v
o
(1  w}p
Epp=
o
o
(1l226a)
o o
Fpp = (1l226b)
(1 w)p
o
and the viscous flux vector Fvp is
(11227)
o
The inviscid and viscous Jacobian matrices are:
82
Chapter 11



~
00 ''
Q)IQ)
..o1~1
II
I I
~ ~~
 ,. ....
"" c
s:: e
+ +
"..,.. :f?'
0:
e o:::~
......... ..... ....
+ +
s: "'" +
~ I
I
I I
I I
il
..,
..:::::::.1
I
t~ ... +
I I I
,., ~ I "" I "....
.'! I... I Ii
~ I ' I . ....
I ? ~ I I I + ~ ..
~ ; ~I I ~ IrE:
s:: e .. I ~ .....
+~~ ~ ~Ll
~  ~ ~~ I
.s:...._.. s:: ~: I
s:: I ,
t +
I "..,.. I
I~~ ~r +i,1
0' + ~t  +.... ~ I I ~
s:: .:::... .'! s::...... I
+ ~ .'! ~ I I
~ ~I':::~t ~~I
~ ~1~ ;~I .
tI""Yot ,... ,, ......
~ ~ .'!
+ I I I 0
 ..
......... ....
~ "'"
~ I
 .........
.....
..... ....
+
+
Transformation of the Equations of Fluid Motion
83
I
I I
I 11 I,;>, I
I ~ I::' I
I:=:; I ~ I
',I I I I ~
~ I ~ I '"
o ,..._.. I "' I ~
1T~
I
+
(:3 ~
~ ~ Ii .<: ~
;>,"'_" ..._.. ~ ;;;..._..
e I $ + (:: I
·r~r
I~
I ~
I + I '"
~ , ~
~ I .....
+ ,
~ (:3 I ~
~ ~:£.. , ~IC'l
~ + ~ ,I
I (:3 I Q)
~ ~ J:.: I.i::. J:.:
~ ~ + I I ~ I
~~4
I ~
I ~
I ~
I + +
I ;; "'(:3
J ~ ::=:::
I + ~
I ;:l
I ~ ~
I '' ~~ I ;:l
o I I
I
+
...
~
+
84
Chapter 11
.

I
N \,oJ
o

I
+ 1 I
__ ~ I
':l"" ,... "I
1 7' I
1 II
~I~
~ ::PI~ I
...___" 2"' 1
...___" I
I I I I I
I
rt+
+ ...... :i"' S' 0
il" S" ",......
...... 1 "'1;:)1c...., "'1;:)1c....,
c...., I ....__..,.,......___...
"I::l I <:! :P ~.:I .:I
....__..,.,
.:I ~
'"
o
+ s:
+ S
f.J.J.
+ s: s: I s: 0
............ ...... I ...
:l"' c...., "I::l I c...., I "'1;:) I c....,
1 ~ ;'1;'
~I~ .:I I
~ I
<c::« I
...... I
c...., I
"I::l lei ..___.. I
'...:jt1
"'VI~ I 0 I 0 I c>
'"'I I I I
~ I I
...1 I
"'1;:) I c...., II I I
;' I I
I I I
Transformation of the Equations of Fluid Motion
85
11.8 Incompressible NavierStokes Equations
As introduced in Chapter 8, a common scheme to solve numerically the incompressible NavierStokes equations is the modification of the continuity equation to include an artificial compresibility term. If T is used to denote this artificial compressibility and fJ is used to represent its inverse, then the incompressible NavierStokes equations in dimensional form are given by
8Q 8E 8F eo 8Ev 8Fv eo; +++=++
at 8x 8y Bz 8x . 8y Bz
(11231)
where
(11232)
(11233)
s; = [~:l
T,;z
(11234)
[f3V
F vu
 v2+p
VW
(11235)
(11236)
[f3W 1
G= wu
:~+p
(11237)
Gil = [~%%l
Tz!/
Tu
(11238)
and
2v8u 8x
T.,  T .. = V (~ + ~:)
_ 2v8u
T!/1I ay
86
Chapter 11
TIZ  TZI = II (8W + 811.)
8x {Jz
_ 2118w
Tu 8z
Equation (11231) may be expressed in a nondimensional form if the variables are nondimensionalized according to nondimensional terms defined previously in Section 8.2.1. The nondimensional form of the incompressible NavierStokes equation in a flux vector form is:
where
Q' = [i~l
[~'u' I E' = [~~ I
E. = u·2 + p. (11240)
u·v· v r..
%11
u·w· •
T:u
[~.v. I F: = [~l
v·u· (11242)
F·
 v· + p.
v·w· TIIZ
[~w. I c: = [~~ I
w·u· (11244)
C·=
w·v· v T.
w·2 + p. zy
•
Tn
where
2 Bu:
•
T = 
II Reoo 8x. (11239)
(11241)
(11243)
(11245)
(11246)