This action might not be possible to undo. Are you sure you want to continue?
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS by
Kang Yul Huh and Michael W. Golay Energy Laboratory Report No. MITEL83011 August 1983
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS
by
Kang Yul and
Huh
Michael W. Golav
Energy Laboratory and Department of Nuclear Engineering Massachusetts Institute of Technology Cambridge, MA 02139
MITEL83011
August
1983
Sponsored by: Boston Edison Co. Duke Power Co. Northeast Utilities Service Corp. Public Service Electric and Gas Co. of New Jersey
2
ABSTRACT
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS by Kang Y. Huh and Michael W. Golay
A computer code is developed to predict the behavior of the hydrogen gas in the containment after a lossofThe conservation equations for the four coolant accident. components, i.e., air, hydrogen, steam and water, are set up and solved numerically by decoupling the continuity and momentum equations from the energy, mass diffusion and turThe homogeneous mixture form is used bulence equations. for the momentum and energy equations and the steam and liquid droplets are assumed to be in the saturation state. There are two diffusion processes, molecular and turMolebulent, which should be modelled in different ways. cular diffusion is modelled by Wilke's formula for the multicomponent gas diffusion, where the diffusion constants are dependent on the relative concentrations. Turbulent diffusion is basically modelled by the ke model with modifications for low Reynolds number flow effects. Numerical diffusion is eliminated by a corrective scheme which is based on accurate prediction of crossflow diffusion. The corrective scheme in a fully explicit treatment is both conservative and stable, therefore can be used in long transient calculations. The corrective scheme allows relatively large mesh sizes without introducing the false diffusion and the time step size of the same order of magnitude as the Courant limit may be used.
A
3
ACKNOWLEDGEMENTS
This research was conducted under the sponsorship of Boston Edison Co., Duke Power Co., Northeast Utilities
Service Corp. and Public Service Electric and Gas Co. of New Jersey. support. The authors gratefully acknowledge this thank all those who assisted them
The authors
during this research especially Dr. Vincent P. Manno, Mr. B.K. Riggs, Ms. Rachel Morton and Prof. Andrei Schor. work was expertly typed by Mr. Ted Brand and Ms. Hakala. Eva The
5 SMAC Scheme and Compressibility CHAPTER 3: DIFFUSION MODELLING 3.3 Tensor Viscosity Method Corrective Scheme by Huh 4.1 Mesh Point Implementation Mesh Interface Implementation 4.3.1 Continuity and Mass Diffusion Equations 2.4.1 Skew Differencing by Huh 4.2.4.3.4.4 ADI Schemes Conservation 4.3.4.2.1.2.1.1 Derivation of Turbulence Equations 3.3.2 Turbuletit Diffusion 3.2 .1 Problem Statement 2.3.4 Solution Scheme 2.2.4 4.1 4.2 Comparison of Explicit.2 Limitations 2.4TABLE OF CONTENTS Page ABSTRACT ACKNOWLEDGEMENTS TABLE OF CONTENTS LIST OF FIGURES LIST OF TABLES NOMENCLATURE CHAPTER 1: CHAPTER 2: INTRODUCTION PROBLEM STATEMENT AND SOLUTION SCHEME 2 3 4 6 12 13 16 21 21 23 25 27 28 29 29 30 30 37 37 42 43 44 45 48 48 53 57 59 62 76 79 81 82 82 83 84 86 87 88 89 91 2.3 Low Reynolds Number Flow CHAPTER 4: NUMERICAL DIFFUSION Truncation Error Diffusion 4.2 Governing Equations 2.1 4.1 Assumptions 2.3 Truncation Error Diffusion in a Two Dimensional Problem 4.3 Discussion of Skew Differencing and Corrective Schemes Raithby's Scheme 4.1 Molecular Diffusion 3.1.4.2 Physical Constraint 4. Implicit and 4.3.3.1 Truncation Error Diffusion in a One Dimensional Problem Another Approach for a One Dimensional 4.3 Basic Assumptions and Limitations 2.3.2 Crossflow Diffusion 4.2.2 Energy Equation 2.1 Diffusion Convection 4.2 Truncation Error Diffusion 4.2.2 ke Model 3.4.2 4.
5
Page 4.4.3 Accuracy 4.4.4 Stability 4.4.4.1 Explicit Scheme 4.4.4.1.1 One Dimensional Central Differencing of Convection Term 4.4.4.1.2 'One Dimensional Donor Cell Differencing of Convection Term 4.4.4.1.3 Two and Three Dimensional Donor Cell Differencing of Convection Term 4.4.4.1.4 Stability Condition in Terms of Cell Reynolds Number 4.4.4.2 Implicit Scheme 4.4.4.3 ADI Scheme CHAPTER 5: RESULTS . . . . . . . . . . . . . . . . . 91 93 94 94 98 101 106 113 117 119 119 119 123 123 125 126 126 132 132 138 158 164 164 164 . . . . 168 168 168 .. . . . 170 176 181 184 204
5.1 Natural Convection Results 5.1.1 Updating of the Reference State 5.1.2 Energy Convection 5.1.3 Heat Transfer Modelling 5.1.4 Turbulence Modelling 5.2 Numerical Diffusion 5.2.1 Truncation Error and Crossflow Diffusion 5.2.2 Validation of Huh's Correction Formula 5.2.3 Skew Differencing Scheme 5.2.4 Corrective Scheme 5.3 ADI Scheme CHAPTER 6: 6.1 6.2 CONCLUSION . . . . . . . . . . . . . . .
Physical Models Numerical Schemes RECOMMENDATIONS FOR FUTURE WORK .
CHAPTER 7: 7.1 7.2
Physical Models Numerical Schemes . . . . . . . . . . . . . . . .
REFERENCES APPENDIX A: APPENDIX B: APPENDIX C: APPENDIX D:
ANALYTICAL SOLUTION FOR THE PROBLEM IN FIGURE 5.3(B) . ............ ADI SCHEME.FOR MOMENTUM EQUATION . . . . . . . . . . . . . . . COMPUTER PROGRAMS TYPICAL INPUT FILES ...........
6
LIST OF FIGURES Number 4.1 4.2 Illustration of the crossflow diffusion in a single mesh with pure convection ... ... Illustration of the crossflow diffusion in a row of meshes aligned in the xdirection with pure convection . ........ Geometry for explanation of the crossflow diffusion in a two dimensional . . . Cartesian coordinate ...... . Page 63
65
4.3
...
.
.
66
4.4
Geometry for explanation of the crossflow diffusion in a rotated two dimensional Cartesian coordinate . . . . . . . ....... Comparison of the prediction formulas for crossflow diffusion constant by De Vahl Davis and Mallinson and Huh . . . . . . . . .
.
.
68
4.5
. . 71
4.6
Geometry for explanation of the crossflow diffusion in a three dimensional Cartesian .. coordinate . ......... ...... A segment of the calculation domain showing grid lines, control volume and notations in Raithby's scheme . ........ Three possible flow configurations for mesh interface implementation of the corrective scheme in a two dimensional ....... .. ..... . . ... problem ...
.
72
4.7
. 80
4.8
.
85 92
4.9 4.10
Profile assumptions of 0 in space and time for various differencing schemes . ....... Three distinct shapes of the parabola, f(t), according to the coefficient of second order term ................. Stability condition in the plane (C xd x ) for an explicit scheme with central differencing of convection term in a one .......... dimensional problem . ... Stability condition in the plane (C ,d) for an explicit sc'heme with donor cell differencing of convection term in a one dimensional problem ..............
. . 97
4.11
.
. 99
4.12
.102
7
4.13
The region in the complex plane where (al+a2 ) should exist for the stability of an explicit scheme with donor cell differencing of convection term in a general two dimensional problem .............. The region in the complex plane where the amplification factor, r, should exist for the stability of an explicit scheme with donor cell differencing of convection . term in a general two dimensional probl Stability condition in the planes (C ,d ) and (C ,d ) for an explicit scheme wfthx donor elY differencing of convection term in a general two dimensional problem .
.104
4.14
. . . 105
4.15
.
. .107
4.16
Stability condition in the plane (C ,d ) etc. for an explicit scheme with doiorx cell differencing of convection term in a general three dimensional problem . ..... . .108 Stability condition in the plane (Rx,C ) and (Ry,Cy ) for an explicit scheme with donor cell differencing of convection term in a general two dimensional problem . .. .111
4.17
4.18
Stability condition in the plane (Rx Cx) for an explicit scheme with donor cell differencing of convection term in a two dimensional problem with f=v/ u/x.... .. 112
4.19
Stability condition in the plane (Cx , d x ) for an implicit scheme with donor cell differencing of convection term in a one dimensional problem .............. Stability condition in the plane (R xCx) for an implicit scheme with donor cell differencing of convection term in a one ........... dimensional problem . ... Natural convection flow field due to heat transfer between the obstacle and air in the containment at t=10.8 sec . .
.115
4.20
.116
5.1
.
. .
. 120
8
5.2
Natural convection flow field due to heat transfer between the obstacle and air in the containment at t=62.5 sec . ..... Problem geometry for the two cases, (A) parallel flow and (B) cross flow, for evaluation of numerical diffusion . . . Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 10x10 meshes along the line CD in Fig. 5.3(A) . Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CD in Fig. 5.3(A) . Comparison of the analytic solution, analytic solution with increased diffusion constant and numerical solution with donor cell differencing of convection term in 10xlO meshes along . the line CA in Fig. 5.3(B) ........ Comparison of the analytic solution, analytic solution with increased diffusion constant and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CA in Fig. 5.3(B) . . . . . . . . .... Problem geometry with arbitrary angles of e and 81 to compare De Vahl Davis and Mallinson's and Huh's formulas for prediction of crossflow diffusion constant . . . . . . . . . . . . . . . . . Comparison of the numerical solution with donor cell differencing of convection term and analytic solution with increased diffusion constant, (D+Dcf), by Huh's formula along the line BD'in Fig. 5.8 ... . . . . . . . . . . . .
121
5.3
127
5.4
128
5.5
.129
5.6
. ..
130
5.7
. 131
5.8
. . . 133
5.9
134
CA in Fig. . . . . . . numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig.16 143 5. 137 5. . . skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with 6 = 63.. . 5. .13 139 5. .10 Comparison of the numerical solution with donor cell differencing of convection term and analytic solution with increased diffusion constant. Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 6x6 meshes along the line . . 5.convection problem with 6 = 450 Comparison of the true solution and solutions by donor cell scheme. . problem with 6 = 63.. .3(B) . . . ..8 . 144 .3(B) . . (D+DDM).9 5. .430 . Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 10x10 meshes along the line CA in Fig. .. .17 Comparison of the analytic solution. . 135 5.3(B) . .. . True solution and donor cell solutions with and without diffusion correction for a pure convection . . 5..15 . by De Vahl Davis and Mallinson's formula along the line BD in Fig.43* . ..14 140 5. . . . 142 5. .. 5..11 Comparison of the true solution and solutions by donor cell scheme. .. .12 . skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure . Donor cell solutions with various diffusion corrections for a pure convection problem with 8 = 450 . 136 5.
5. .. .23 .20 . numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig.. ... Recirculating flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme Detailed flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and riesh interface implementations of the corrective scheme Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme . ... Comparison of the true solution. 5.. Diffusion constant corrections at each interface for the mesh point and mesh interface implementations of the corrective scheme for recirculating flow field given in Fig.19 ... 145 5. 153 .. . . ... . . . 152 5. .. . . . .. .147 5. 151 5...18 Comparison of the analytic solution.3(B) .25 Diffusion constant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .22 150 5. 148 5.. 5. .. . . .... . .. . .24 .21 S. . .. . . . ... . 146 5. . Simple recirculating flow field for test of the implementation strategies of the corrective scheme .19 .. donor cell solution and solutions by the mesh point and mesh interface implementations of the corrective scheme for the recirculating flow field given in pig. .19 . ..... . . .10 5. . . .. . .
. . .0 sec along the line AB in Fig. .30 5. . .. . . 1 178 . . 5.. .cx T where 4 is given by.5 sec and the Courant limit of 1. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. . .30 for At = 1..1I 5.. . . .0 sec and the Courant limit of 1.. .3(B) = e. .. with donor cell differencing of convection term without any diffusion correction .3 sec. . .3 sec. 156 5. . 0. .. . .0 sec . . . . . . .30 for At = 0.. 0. .30 with . 5. with mesh interface implementation of the . 163 A.. . . ..3 sec. with mesh point implementation of the corrective scheme .. 5.. 5. . Steam concentration distribution after one time step.. 0. 157 159 5.26 Diffusion constant corrections of the mesh interface implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment ..31 1i60 5. 154 5. different time step sizes Boundary conditions in terms of 4 for the problem in Fig. . . Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. .0 sec .33 Comparison of the ADI solutions for the profile of 4 at t = 10. . . .32 161 5. . Steam concentration distribution after one time step. . .28 . corrective scheme . 155 5. .. Problem geometry for comparison of the explicit and ADI schemes . .29 Steam concentration distribution after one time step. . .27 .. .
1981 . . . .1 List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research. 78 4.12 LIST OF TABLES Number 4. . . . . . . . Comparison of the Von Neumann analysis and numerical experiments for the stability condition in terms of the cell Reynolds number in an explicit. . . . .114 . . . . . . Page . . .2 . donor cell differencing of the convection term . . . 2D. . . . .
D y.y and z directions Diffusion numbers in x and y directions Internal energy Function of Body force vector per unit mass Drag force vector per unit mass Gravity Grashof number Enthalpy Heat transfer coefficient Imaginary unit.D d .d xy e f Gr h I Im k g M Nu P Pe p y Az Pr Q Prandtl number Heat source ./Imaginary (1) Turbulent kinetic energy (2) Thermal conductivity (3) Ratio of specific heats.13 NOMENCLATURE Letter A c c Definition Area Specific heat at constant pressure Specific heat at constant volume Courant numbers in x and y directions (1) Dimension (2) Diffusion constant (3) Divergence (4) Discriminant of a quadratic equation Binary mixture diffusion constant between air and hydrogen Binary mixture diffusion constant between air and steam Binary mixture diffusion constant between hydrogen and steam Crossflow diffusion constants in x.C x y D D ah D as Dhs D . cp/c Length Mass Molecular weight Nusselt number Peclet number Effective Peclet number (1) Pressure u u (2) L* ax or Ax Ay Ax + p v C .
1 v u Mesh configuration. tan.14 Letter R Re Re Rt Rx S Sc T t At u v v w Ax y Ay Az Greek Definition Universal gas constant Reynolds number Real Turbulence Reynolds number Reynolds number in x direction Source Schmidt number Temperature (1) Time (2) Parameter of a quadratic equation in Von Neumann analysis Time step size xdirection velocity ydirection velocity Velocity vector zdirection velocity xdirection mesh spacing Mole fraction in multicomponent gas ydirection mesh spacing zdirection mesh spacing a ala2 8 6ij C p v 4(1) 6 61 Thermal diffusivity or diffusion constant Parameters used in Von Neumann analysis Thermal volumetric expansion coefficient Kronecker delta Turbulent dissipation rate Density Kinematic viscosity Vpiscosity Phase change (2) Any general conserved quantity Velocity direction. 1 tan STime a Z Superscript (1) Vorticity (2) Amplification factor in Von Neunann analysis constant of heat transfer Traction tensor due to viscosity Summation Time step n .
15 Subscript cell cf DM f g i=l 2 3 4 L t m mix new ob old R t total US Cell Crossflow diffusion De Vahl Davis and Mallinson's Fluid Gas Air Hydrogen Steam Liquid Droplets Left hand side Liquid Molecular Multicomponent gas mixture New time step Obstacle Old time step Right hand side Turbulent Total calculation domain Upstream .
The four components. Some simplifying assumptions are made concerning the thermodynamic state in the slow mixing stage.I~~U1i* 11*11044~1 1~. steam and liquid droplets are assumed to be in thermodynamic equilibrium and the relative humidity is assumed to be 100%. The governing conservation equations are decoupled in order to simplify the solution procedure. hydrogen. The research work reported here is primarily con cerned with the formulation and validation of the physical models and numerical schemes in the second slow mixing stage. From the hydrogen transport point of view. The error due to decoupling is negligible in a slow transient where the .__ 16 CHAPTER 1 INTRODUCTION One of the major concerns in the Three Mile Island (TMI) accident was hydrogen gas accumulation in the containment. it is necessary to understand such problems. The distinct feature of the second stage is a much longer time scale in comparison with the first blowdown stage. the response of the containment during an accident can be divided into two stages. the fluid dynamic phenomena in the containment in order to justify those remedies. the fast blowdown stage and the slow mixing stage. although it may be less than 100% when there is no liquid component present. Some remedies have been proposed to deal with However.( 1~111111_~. air.
Finally. Diffusion occurs by two independent mechanisms.17 state change over one time step is small. are determined by their own transport equations. In the second step the energy and mass diffusion equations without phase change and turbulence equations are solved using the flow field obtained in the first step. k. Convection is assumed to occur as a homogeneous mixture. The turbulent diffusion constant is predicted by the kc model [45). c. The molecular diffusion constant is predicted by Wilke's formula [73) for multicomponent diffusion and ChapmanEnskog formula [5) for binary diffusion. and the total diffusion constant is the sum of the diffusion constants of those two mechanisms. resulting in the same convection velocity for the four components. The diffusion constant. the phase change is taken into consideration to maintain 100% relative humidity. and turbulent dissipation rate. The turbulent kinetic energy. In the first step the continuity and momentum equations are decoupled from the energy and other scalar transport equations and solved by the Simplified Marker and Cell (SMAC) method in order to obtain the flow field. Convection and diffusion are the central issues in physical modelling efforts of hydrogen transport. molecular and turbulent. . of each component in multicomponent gas depends on the mole fraction of that component. The turbulent kinematic viscosity.
Since the error usually appears as an additional diffusion. truncation error diffusion and crossflow diffusion. The leading issue in numerical modelling of convection and diffusion is to minimize the error that occurs in the numerical solution procedures.18 which is the diffusion constant for momentum transport can be calculated directly from k and E. The numerical diffusion has two different sources. has been used to describe the numerical error in general. Truncation error diffusion is a one dimensional profile error and crossflow diffusion is a multidimensional operator error [68) of the finite difference equation. The turbulent Prandtl and Schmidt numbers are assumed to be equal to one. There has been much debate on the numerical diffusion and many schemes have been suggested for the past two . Therefore. the major obstacle in accurate numerical modelling of convection and diffusion has been the crossflow diffusion error which arises in donor cell treatment of the convection term in multidimensional problems. the term. This is because the gradient of the scalar quantity under consideration is small in the flow direction in comparison with that in the direction normal to the flow. numerical diffusion. The effective diffusion constants of the two errors are of the same order of magnitude. Truncation error diffusion occurs in the flow direction while crossflow diffusion occurs primarily in the direction normal to the flow. however the latter turns out to be the dominant error source in most convection dominant problems.
The corrective scheme is inherently better than the skew differencing scheme in that it is conservative and does not affect the simple solution procedure. The conservative property is essential in a long transient problem like the hydrogen transport in the containment. Chang's methods [16) are examples of skew differencing schemes and Huh's corrective scheme and tensor viscosity method [25] are examples of corrective schemes. ADI and implicit schemes. have been tested for recirculating flow problems. The schemes presently being used can be divided into two categories. although with different stability conditions. a recent review paper on this topic [65) shows that there is no scheme universally acceptable consistent with using reasonable mesh spacings and computation times. Two implementation strategies for the corrective scheme.19 decades. This situation is partly due to misunder standing of the sources of numerical diffusion and also partly due to use of inappropriate approaches for its elimination. skew differencing and corrective schemes. Raithby's [53) and S. The stability conditions for each of the aforementioned schemes can be obtained by a Von Neumann analysis and turns out to be consistent with numerical experiments. However. The corrective scheme can be implemented with any of the explicit. The mesh interface . mesh point and mesh interface implementations.
. [49) includes all the Development Laboratory) [6]. The suggested physical models and numerical schemes have been used to simulate the LOCA experiments performed in BattelleFrankfurt [44) and HEDL (Hanford Engineering Ref.20 implementation has always given physically reasonable solutions and may be used extensively for all diffusionconvection problems. simulation results compared with experimental data.
for dealing with this problem... The hydrogen is generated by radiolysis and chemical reaction between water and zirconium in the cladding and may react explosively with oxygen in the air. a large amount of steam and water will come into the containment in the fast blowdown stage increasing the containment pressure. The major safety concern is that of keeping the containment pressure below a certain level to prevent a large scale leakage of radioactive materials.21 CHAPTER 2 PROBLEM STATEMENT AND SOLUTION SCHEME 2. In addition to the pressure increase due to the primary coolant. containment inerting. hydrogen generation gave a serious concern about the integrity of the containment in the TMI2 accident. it is essential to understand the fluid dynamic phenomena in the containment. e.g. After a while the slow mixing stage follows the initial blowdown stage and continues for an extended period of time. the hydrogen has received much attention in the safety analysis of nuclear power plants. In order to justify the design of any mitigation system. installation of ignition devices. Thereafter.1 Problem Statement Once a lossofcoolant accident (LOCA) occurs in a Light Water Reactor (LWR). use of flame suppressants and enhanced venting capability. There have been some mitigation procedures suggested. .
laminar and turbulent flows. some simplifying assumptions should be made to use the numerical procedure without impairing the acceptable solution accuracy._ __ _ __ _ I_ ~ 22 Since a numerical method is suitable for this purpose. The most difficult aspects of this analysis are the complicated geometry and chaotic postLOCA conditions. phase change and heat transfer between the gas components and wall. a good computational tool has been required to predict the hydrogen concentration distribution. e. . thermal nonequilibrium. etc. Therefore.g..
t = VD.T) (2.6) Equation of state: p = f p (ii .4) Turbulence kinetic energy: Dk 1 t 3k 3 P 1 k ok x . Continuity: V'v= 0 pl + Vvv ] = Vp + Vv = 0 (2.xk k 2 v t BT C2 + kg k Pr 3x 2k k t (2.2) Energy: p[e e + Vvh] + = VkVT + e fg (2.23 2.1) Momentum: pgI + d + V0 (2.Vp i + 4. The physical implications of the conservation equations will be given with their basic assumptions and limitations. (2. I+ 3x 1 .S + + t Bui P au k aui + ax~ ) x k X axk gk Pr t t 'T xk (2.5) Turbulence dissipation rate: D 1 _ 1t StE xk Dt P Txk 0E ] + C t*( 1 p k ui xk ouk au.2 Governing Equations The governing conservation equations are set up to describe the postLOCA fluid dynamics in the containment.7) .3) Mas s diffusion: S+ V*vp i .
The subscript i denotes the four components as . .24 Constitutive equations: 'N = f (p . = f (Pi.T) (2.11) k = .18) p = Epi (2.17) t = C k 2/C E (2.T) (2.t (2.10) v = vm + Vt (2.12) (2.14) m = fk (PiT) (2.8) f (v) (2. Im + Dit it (2.19) Note that 1.k + k t m Dim = fD(Pi'T) (2.16) kt = .15) Dt = v t (2.13) D.9) D. = D.
u.k C for a three dimensional case. (2.22) There are eleven conservation equations. The turbulence equations are written in terms of the Cartesian components in tensor notation.2.1 Continuity and Mass Diffusion Equations The continuity equation is redundant with the four mass diffusion equations and there is'an inconsistency between them. H2' steam. The mass diffusion equations are summed up for the four components.21) D4 = 0 (2.20) i = 2 = 0 43 + 04 = 0 (2. liquid droplet. Therefore. there is one moreequation than is required. air. p. i=14. v. . T. three momentum. four mass diffusion and two turbulence equations with ten primary unknowns. w.i (4). The phase change occurs between steam and liquid droplets and the diffusion constant of liquid droplets is equal to zero. one continuity. one energy.25 follows: i=l i=2 i=3 i=4 2. 2. 3.
27) Consequently Eq. 2. 2. 2.26 . 2. 2.27. it is reasonable to assume that Eq.26 can be reduced to the following Eq.(Ep i ) a .Vp.23) The sum of the phase change terms is equal to zero.26) It is a reasonable assumption in a turbulent flow regime that the four components have the same diffusion constant.25 and Eq. 2. at p= 0 (2.27 remain valid throughout the transient. Since the concern is only for the slow mixing stage after a LOCA.25 and Eq.4 are consistent only if the temporal and spatial variations of the total density are very small.) + V(Ep Jt i = (V'D. ap at S+ Vp = E(VDiVp i ) 1 (2.1 and Eq. .25) Vp i = 0 (2. 2.24) Now it can be seen that the continuity equation is valid if and only if the following Eq. Therefore.26 are satisfied. Vp = 0 (2. 2.) i + E 1 (2. Eq.
29) The second term on the left hand side of Eq.29 denotes the energy change due to the concentration change in a given control volume.30 is substituted in Eq.iei) + V'[Ep. I at t + V [vp.31) The thermal equilibrium assumption is incorporated in Eq. [c pihi . 2.3. (p.i + V. .29 to give the following. It is clear that the major contribution to this term will come from the latent heat of the phase change. is also of an approximate nature and the exact form is given in the following.27 2.28) The temporal term can be divided as follows. 2.2 Energy Equation The energy conservation equation.30) Eq. = VkVT (2. E it p + e. 2. ae.i = V'kVT (2.ih i ] = V*kVT (2. (2. Eq. 2.2.31 so that the internal energy and enthalpy can be written in terms of the temperature. Be. LeaPi LeQ.2. p + £e.ih.
T h. 2.34) 2.1 = c Vl. It is necessary to clarify these assumptions and their limitations for a safe use of the given physical models and solution scheme.28 (Picvi )~ +V"[vTZpic pi] + efg 9kVT = (2.c . . 2.3 and Eq.32 become identical by defining average internal energy and enthalpy as follows.32) where ei i = eg4g + e 4) Sefg g e.T 1 Eq.T P P e Piei p ih p.33) h (2. P (2 33) (2.3 Basic Assumptions and Limitations The governing equations are based on some simplifying assumptions about the physical phenomena in the containment after a LOCA. Pic vi T c p p. = c p1 .
3. 2. 5. with no slip. The phase change rate should be moderately small because the solution scheme decouples the phase change term from the energy conservation equation. 3. The relative humidity may be less than 100% when there is no liquid component left. 3. The four components are in a state of thermodynamic equilibrium. . 4. The temporal and spatial variations of the total density should be negligibly small because of the incompressibility assumption.2 Limitations 1. Dt and k t over one time step should be small because the turbulence equations are decoupled from the velocity field calculations. The change of the diffusion constants. The turbulent Prandtl and Schmidt numbers are equal to one. The four components are treated as being nearly incompressible and the phase change rate is moderately small.3.29 2. The relative humidity in the containment is 100%.1 Assumptions 1. 2. 2. The four components all have the same convection The liquid droplets move with the gas components velocity. vt.
Pi(4). In the first step the SMAC scheme is used to get the convergent velocity field with zero divergence for every mesh. v.5 SMAC Scheme and Compressibility The SMAC (Simplified Marker And Cell) scheme [21J has been used for the incompressible fluid flow with Boussinesq approximation for the buoyancy force. The equation of state.1Eq. k. p.i Since it is too difficult to solve the whole system of the equations and obtain a consistent solution for all the unknowns. 2. When there is no liquid component left. .4 Solution Scheme Eq.30 2. It solves the . is used to update the reference pressure and reference density which is important in calculating the buoyancy force.7.7 are a coupled set of equations with primary unknowns. Eq. continuity/momentum equations set and other scalar transport equations set. the conservation equations are decoupled into two sets. energy equation and two turbulence equations. which are the four mass diffusion equations. 2. 2. the relative humidity may be less than 100%. T. u. The phase change is assumed to be zero in this step. the obtained velocity field is substituted in the scalar transport equations. The SMAC scheme is In the second step.5. explained in detail in section 2. E. . In the third step phase change is taken into consideration to maintain 100% relative humidity. 2.
36 and Eq. Since the continuity equation used in the SMAC is an incompressible form. which can arise with net inflow or outflow boundary conditions.36) .37)y (2.1_ _~ 31 continuity and momentum equations to get the velocity and pressure field in an iterative way.35) xdirection momentum: au au + u+v at ax au ay 1 p ax + Pgx + a 2u 2 ax 2 2au + 2 2 (2.37 as follows.37): .36) ydirection momentum: av av ua av 1 + pg + a y x2 Nv 2 (2.ax 2. it is possible to modify the SMAC scheme to accommodate a slight compressibility effect.. Continuity: au ax av ay 0 (2. 2. the divergence of the velocity field vanishes everywhere. (Eq.e. 2. ay (Eq. The continuity and momentum equations are given in two dimensional Cartesian coordinates in the following. the SMAC scheme can be used only for incompressible flow calculations. However.37) Py The vorticity conservation equation is derived from Eq. i. 2.
ax 2 2.37 as follows.+ y ax ay The basic idea of the SMAC scheme is to separate each calculational cycle into two parts.32 a C 2 a2 + u+ V = v+ ) at .39) where av au D= . u n+lu n u At n+l At n n 1 Apn+l A + n 1 p Ax n+l + gn 1 Ap p by (2. ay 2. 2. 2.37): a2 = 2 ax 22 axay a 2 ay 2 (a2D + a2D aD +1 2 at Re ax 2 ay (2.37 as follows.41) where fn and gn are the explicit quantities at time step n.36 and Eq. . (Eq.36) + ~(Eq. 2.x ay ax 2 ay2 where au av (2.38) =ay ax The Poisson equation for the pressure field is also derived from Eq.36 and Eq.40) S (2. 2. In the tilde phase the approximate velocity field at time step (n+l) is obtained from Eq. the socalled tilde phase and pressure iteration.
2.33 Since the pressure field in Eq. 2. The equation of vorticity transport.42) av .v) D (2.43) ax ay It can be seen from Eq. the following relations hold after the tilde phase.41 have the right vorticity.44) Eq. The impor tant point is that un + l and vn+l calculated from Eq.39 may be rewritten as A finite difference form of Eq. 2.40 and Eq. 2. 2.41 are still unknown. 2. V2p = f(u.45) . shows that the vorticity is independent of the pressure field. Eq. a guessed pressure field or the pressure field at time step n is used to start the iteration.44 is given in the following. 2. 2. _ Pi+j2P +P iPij+ pij f(uv)  Dn + l D n D D Ax 2y t (2.39 that the pressure field is not right due to the fact that the divergence of the velocity field is not equal to zero. au ay au av ax  (2. Therefore.40 and Eq. follows.38.O 0 (2.
49) .j) 26p( 1 Abx 6p = h + _) Ay 1 At( A + 1 Ay 1 Dt D (2. no more correction will be made. (6u) L =  pAx (6u) R = t6p pAx (2. S+ at V"(p) = 0 (2. 2. Once the zero divergence velocity field is obtained.. while the continuity equation has the following compressible form.36 and Eq. sidering the cell (i.46) (2.47) ) The pressure correction is given in terms of the residual divergence for every mesh. The slight compressibility means that the NavierStokes equations may be solved safely with the incompressibility assumption.34 The pressure correction formula is obtained by conn+l only and putting D. The velocity correction formula is derived from Eq. 2. equal to zero.48) The SMAC scheme can be extended to a slightly compressible fluid flow.37 so that it does not change the vorticity implemented in the tilde phase.
2.35 From Eq.p_ p at 1 min/V At n p n n (2. : in n =Vp v  total net inflow of mass at the boundary during the time At.50 may be given from the previous time information and boundary conditions.51) v p where m.uniform over the whole containment. The associated physical assumption in Eq.52) (2.49 the nonzero divergence may be derived as follows. 2. 1 . Now the problem is how to converge to a predetermined nonzero divergence field. This can be done with the following modification of the pressure correction formula of the SMAC scheme. V: total volume of the containment. D = V'v 1 p p at V p (2.52 is that the pressure perturbation due to the inflow at the boundary propagates throughout the containment instantaneously and the pressure and density increase is. . 2. 2.50) The two terms on the right hand side of Eq.51 and Eq.
50. The velocity correction formula remains the same as Eq.(DD 1 1 + ) 2At( Ax Ay (2.6p = . 2. .48.53) where D O is given by Eq. 2.
Although turbulent diffusion is greater than molecular diffusion by a few orders of magnitude.37 CHAPTER 3 DIFFUSION MODELLING The hydrogen. Diffusion occurs by two different mechanisms. in the containment after a LOCA. It gives the diffusion constant of each component in terms of . air and steam. molecular and turbulent. air. 3. the diffusion constant of each component is calculated separately by Wi~k's formula [73] in Eq. proper modelling of the latter is important because the molecular diffusion of hydrogen is significantly greater than those of other gases and also because there may be a laminar flow region in the containment.1. steam and liquid components are transported by convection and diffusion in the containment. the process Since there may be three gas of diffusion is also increased.1 Molecular Diffusion Molecular diffusion occurs by the collisions of gas molecules. components. Convection occurs as a homogeneous mixture resulting in the same convection velocity for the four components. As collisions occur more frequently. which should be modelled independently. The thermal conductivity and viscosity are also diffusion constants in a broader sense for momentum and energy transport. hydrogen. 3.
3. stants are given as follows..38 the binary mixture diffusion constants and mole fraction of that component.2) poAB D.: Binary mixture diffusion constant between the ith and jth components. The property of air will be replaced The resulting diffusion con with that of nitrogen gas.2. The binary mixture diffusion constants are calculated by the ChapmanEnskog formula.: yi: Diffusion constant of the ith component Mole fraction of the ith component D. T3 D = 0. 1Y A B DAB 1Y B yA DBA 1Y C YA CA C YC CB B YC DBC C D = C YB DCB (3.1) where D.AB where DAB: AB Binary mixture diffusion constant in [cm /sec] . Eq.0018583 1 + M A MA MB (3.
In order to calculate the para meters aAB and 0DAB in Eq.3.2. 3. =n [ p n = [+ S n . (3.AB: Dimensionless function of the temperature and intermolecular potential field for one molecule of A and one of B.UIIII ~ II~PI^ICIL1I n~s^  ^ 39 T: p: aAB: MAMB : Temperature in [OK] [atm) Total pressure in LennardJones parameter in [A] Molecular weight QD. The error range of the ChapmanEnskog formula is reported to be about 610% [5).3) CAB = AB For the pair of polar and nonpolar gases a correction factor is introduced to account for the polarity. For nonpolar gases OAB and ODAB are calcu lated by Eq. 3.4) np = np 2 where the correction factor 4 is given by. P n ].lll~*.. OAB = (o A + oB (3. ap = np £ (o n + o p) EE e (3.5) 2 n . the following relations are required.
6947 as . The containment pressure is assumed to be1 atm. 5 = 4. 3.2 E/K = 380 0 K a = 2.9492 x 10 D ah D 1 T .9 x 10 25 25 25 [cm3 3 [cm ] : N2 : an = 17.40 The parameters in Eq.6810 x 10 .9035 = 4.= 1.5 are defined as follows.6 x 10 The fitting functions for the binary mixture diffusion constants are obtained by the leastsquarefit method.6 T1.65 A The polarizability is given for the hydrogen and nitrogen (air) components. a * = a an/ : Reduced polarizability of the nonpolar molecule ~ * = p p/ p co 3: a pp Reduced dipole moment of the polar molecule t * = p //8 p p For the steam component the following values will be used. t* . H2 a n = 7.
8) j=l where ij 8 Mj j ij M The viscosity of nonpolar gases may be obtained by the following [5).10) Viscosity in [poise) Temperature in [K]) LennardJones parameter in [A) Dimensionless nTumber . = n i=l y.lllI_ 41 Dhs = 2.7) k.67 x 10 where : T: a: Svi : o2 2 (3. = 2.8916 x 10 where T: D: [OK1 [cm 2 /sec] 5 T 1.k.lii.__Li~n_~PIIII ~. n mix = Yipi n i E j=l Yj ij (3.6) The viscosity and thermal conductivity can be considered as diffusion constants for momentum and energy transport and calculated similarly as follows [5).8144 (3. i (3.
The ks model sets up the transport equations with some empirical constants for the turbulent kinetic energy.42 The viscosity of steam is. 3. There are several models suggested for Presently the best model calculating turbulence effects. The effective diffusion constant is therefore the sum of the molecular and turbulent diffusion constants.009 cp at 1 atm. Wsteam = 0. and turbulent dissipation rate. k. seems to be the kc model originally developed by Launder and Spalding [45). 200 C There is another formula suggested for the viscosity of multicomponent gas in the following [5].y RT k=l PMiDik k i There is a useful relationship nonpolar polyatomic gases.11) for the Prandtl number of Pr = c (3. E.385 nn y. The .2 Turbulent Diffusion The diffusion process is greatly enhanced by the turbulence of fluid flow.[5].12) where c P is the specific heat per mole at constant pressure. 2 n Yi umix = C 2 i=l Yi + 1. c (3.
+ u. The effect of turbulence appears as the additional terms due to the product of fluctuations which do not necessarily cancel out.14) . 45] The transport equations for the turbulent kinetic energy and turbulent dissipation rate are derived from the continuity.1 Derivation of Turbulence Equations [15. = u.43 turbulent viscosity is given directly in terms of k and C.u.' etc. + Pi Pi = Pi + p.15) .= v ( Dt + ui j u P= Dt ax x 2 k6 ij 3x_) 3 i (3.' j T'= a t x.u.' For example. momentum and energy equations by decomposing the variables into mean and fluctuating parts and averaging the resulting equations.2. 34. These terms are treated by the eddy viscosity concept of Boussinesq and eddy diffusivity concept which are given in the following. u. (3. etc. u. 3.13) (3. 29. .
44 The transport equation for u. 3. 3. A two dimensional cylindrical coordinate form of the kE model is introduced in Eq. The turbulent dissipation rate equation can also be derived in a similar way and some assumptions should be made in modelling various terms that appear in the derivation.c +g t Pr v BT az Tt (rue) + + 1 r ar a(w)= z [) [ pr ar o ar + z a ( a az + C 1 aw [ 2 au 2 t + 2() 2 () p az ar p k 2 2 + 2 2 C + ( r + ) az 2 2 k + k Pr atT zt (3. t + at r a (ruk) + (wk) = [ p r (r oka r ) + (t ) az ok a au aw + ) r aw au 2t + 2( + 1[2() 3 3r p + 2 + u2 2 ] .' u is transformed to the turbulent kinetic energy equation by the contraction.16.2 ke Model The transport equations for the turbulent kinetic energy and turbulent dissipation rate form the basis of the kE model. j=i.2.16) .
3.16 as follows.45 The turbulent viscosity pt is given in terms of k and E in Eq. Since the kE model is applicable to fully .09 C1 1. C 0.17.3 Low Reynolds Number Flow While both turbulent and laminar flows are possible in the containment after a LOCA depending on the hydrogen generation rate and geo.metry.17) The suggested values of the constants are given in the following table. 3.44 C2 1. = C pk 2/ (3. it is difficult to determine whether a region under consideration is in a turbulent or laminar flow. 3.92 ok 1. it ak t a +it m ok lit m a E C The turbulent Prandtl and Schmidt numbers are usually assumed to be equal to one.0 o 1.3 The molecular effects can be included in Eq.
3 exp(R2t ) (3. CP = CP0 exp[2.0 .0 a ui ( 2 p ax ax 2 ) Dk Dt 1 p ak [(.46 turbulent flow. Rt. The model presently being used is just to use the sum of the molecular and turbulent contributions for every region in the containment.5/(l+Rt/50)] C2 =2 where R 2 0 [1.0.18) In the above equations C 1 1 ok and a retain the values assigned in the ordinary kc model. while C1 and C2 are to vary with turbulence Reynolds number. it is required to extend the model to laminar flow regime or to switch to laminar models according to some criterion. The following kc model is a modified form by Jones and Launder [40.2vak 2() k 2 (3.16. Eq. D Dt p ax k [(t Dt+ ) e k + C k i + k x ) au axk C 2 E2 2 t 2. 3. 41).19) = pk2/VE . A more refined model is introduced in this section because consistent treatment of laminar and turbulent flows and transition between them may be required in the future modelling efforts.lit k + l ak aui i + t (ax k + xk auk axi axk e .
47
The C O and C2.
and C 2 0 are fully turbulent values of C
This modified form of the ks model was developed
to cover all the laminar, transitional and fully turbulent regions. The constants were fitted to the data of a low
Re flow in a round pipe to include the effect of laminar wall boundary layer. Although its applicability to a low
Reynolds number flow in general is questionable, the model may be used with some confidence if it has the proper limit of laminar regime with the decay of turbulence. It may also
be used in the transitional regime by assuming an adequate interpolation between the two extremes. Equation 3.18
shows that k and e have the same order of magnitude with the decay of turbulence. In other words, both k2/c and
c2/k will go to zero as k and E go to zero separately. Therefore, the turbulent viscosity Vt which is proportional to k2/E will vanish with the decay of turbulence. In order
to have the proper laminar limit, the total diffusion constant should be expressed as the sum of the turbulent and laminar diffusion constants. It may also be reasonable
to assume that the turbulent Prandtl and Schmidt numbers are equal to one.
48
CHAPTER 4 NUMERICAL DIFFUSION
The final numerical solution involves two types of errors which come from physical modelling and numerical solution procedure. The error in physical modelling is
an intrinsic error which cannot be eliminated by the solution procedure. The error in numerical solution procedure
is the difference between exact and numerical solutions of the governing equation, which usually appears as additional diffusion. It is clarified that there are two
sources of numerical diffusion, truncation error diffusion and crossflow diffusion. This chapter is primarily con
cerned with how to predict and eliminate these additional false diffusions to get an accurate solution.
4.1
Truncation Error Diffusion Truncation error diffusion occurs due to the approxi
mate nature of the finite difference formulations.
The name
of the truncation error originates from the Taylor series expansion where the second and higher order terms are truncated. Although the Taylor series expansibn is not a
proper way of interpreting a finite difference equation, the name of the truncation error will be retained here. Since the truncation error exists in multidimensional problems in the same way as it exists in one dimensional problems, it can be analyzed in the following one dimensional
_I_
I~___
_L~ ~
_
I
conservation equation without any loss of generality.
at where 4: S:
 U +
ax
a
2 ax
+ S
(4.1)
any general conserved quantity source term and diffusion constant a will be assumed
The velocity u constant. [xi, xi+
Equation 4.1 is integrated in the domain
]
and [tn
,
tn+l].
x
Xi
n+1 at dt dx + xi+ x tn+l t n aa 2 a2 ax
x
tn+1 u dt dx tn
dt dx + S Ax At
(4.2)
where At = t n+l t
Ax
n xi_
=
Xi+

When a function f(x) is continuous, there exists a point x=c in [a,b) such that,
b ba
1
f
a
f(x) dx =
f(c)
(4.3)
Using Eq. 4.3, we may transform Eq. 4.2 as follows.
50
t
1
[
n+l
i+f n+f
+
Ax
[ n+ 9 1+h
n+g
i
= 
S[(a4n+g Lx [ ax ) i
(x
n+g ] + s + ax a+ ih
(4.4) (4.4)
where <f <
0 < g< 1 Equation 4.1 may also be reduced to Eq. 4.5. It is fully explicit and the convection term is finitely differenced by the donor scheme.
h+1 At =
n
+
Ax (Q i

1(4.5)
x2 Ilx
n i+
 24P + 4i1 ) + S 1
From Eq. 4.4 and 4.5, obtained as follows.
(4 +)EXACT
and (
)FD can be
n
n
Sn+l i
FD
n i
+
At [ u
i
Ax
bx 2
S n+g _n+g
n
i+1

+
+
l 11 )
(4.6)
+ S]
n+l
i+f EXACT
n
i+f
At
i+h
Ax
i1
+ Ss
+ a( n+g ) i6 + Cx x[( i+6 _ ( ax)n+g ax
]
]
(4.7)
y+b) = f(x. It can be extended to a two dimensional form as follows.y) + (f) af (4.6 from Eq. af f(x+a. 1 < c < n 2 4 + i Ax 2 11 4 2 1 i+d < d < 1 The mean value theorem. 4. we obtain the error at time step (n+l).a i = ( nn+g n+g ax i+a  < a < n i1 8a ( ) n i+b a 2 4n+g ax 2 i+c 2 ox n 1 < b < 0 n+g (a n+g ax i. .9) a + ( ) b AA B some B in the domainB ay whereinterior andpoints denote where A and B denote some interior points in the domain bounded by [x.y+b]. n .3. 4. Eq.x+a) and [y. 4.a(?) ) 2(. n+g 4 i+ Ax n i Ax 1 xji+ n i+l .8) 3 @n + u()n + a( ax i+b ax 4 n+g _ n )n+g .7.i+c T i+d ax where the following relations have already been used. is for a one dimensional case. n+l i+f EXACT n+l i FD n i+f  4P n a n+g i + At [ u( x ) i+a (4.51 Subtracing Eq.
tax . 4. E . 2 n+l (4i ) EXACT  (4+) FD = i n+1 f )2 2fx 1f 2 ) (f 2 2 . The terms A.8 to the following form. . Although the error terms in Eq.10 are based on the assumption of perfect information at time step n. 4. 4.F (4. C and E are truncation error diffusion terms while the terms B. D and F are time derivative terms which become negligible for slow transients. In a stable scheme the error introduced at a certain step decreases in its absolute magnitude in the following steps and most of the error occurs between neighboring time steps.10 are at some appropriate points in the domain [xi .C ax 22 +At g t .9 reduces Eq.10) The derivatives with respect to time and space in Eq. 4.xi+ ] and [tn tn+).A fAxAt axat 2 B utx(a+b) . they may represent the errors over many time steps in a stable scheme.52 Now Eq.D +aAx(cd)Atax 2 +agt .
u ax a22 (4.1. o 1 2 3 L i1 1 i+l N 4(x=0) = 40 N Q(x=L) = NAx = L Two boundary values 0 and QN are given as constant. 4.11 is given as. The analytical solution of Eq. cL (4.12) and C 1 =N cL 1 e C2 ecL 1 . 4= Cecx + C2 where c = u/a .1 Truncation Error Diffusion in a One Dimensional Problem The truncation error diffusion in a one dimensional problem is analyzed for a steadystate case with no source.11) ax Consider a case in which the domain is divided into N equal meshes and the mesh size is Ax.53 4. The results may be extended to a general one dimensional problem if the effects of the transient and source terms are not dominant in determining the profile of 4.
4.15.16) . Equation 4. i+l i = (P+l)( 1 . .15) Equation 4. i+  i = (P+l)(4)  4)i (4.54 Equation 4. (4.14 is the cell Reynolds number or cell Peclet number according to whether the diffusion constant a is the kinematic viscosity or thermal diffusivity.40) .13) where the velocity u is positive. P(4).16 is easily obtained from Eq. i u x i+l2. 4.14) The parameter P in Eq.11 is finitely differenced using donor cell scheme for the convection term in the following.15 and solved for in subsequent procedures.13 can also be solved with the given boundary conditions by reducing it to the following form.14 may be recast to Eq. where P =u a  l i24 + ii (4.i + 4i2 Ax = a (4. 4. Equation 4.
'N1 (1+P) N1  ) 0 N . 4.20 have been obtained without any approximation for Eq. 4. i  0 = 4. 4. 2 1 = (1+P) (4 0) 4' N 92 = N2 (1+P) 2 (  0) = N.17.12 and numerical solution Eq.(1+P) P (1+P) 0 + N .19) (1 .~LI LIII~.13.~l*IIIIX1~~^ ~ ^I ^~Xi 55 Then.20) Now the analytical solution Eq. 1 (1+P) i . Both the analytical and numerical .19 into Eq. we obtain the solution for 4. as. 4. 4.16. (1+P) 1 .LI II .I_.41 = N ) (1+P) [(1+P) N1  2) (4.17) 01 can be expressed in terms of 40 and 4N N N by Eq.1 Inserting Eq.18) (l+P) + 1 Therefore. (1 0 ) is ) P( N ¢0 (1+P) given as follows.11 and Eq.1 (1+P) N 1 (4. 4. 4.
21 and Eq. e (il)Pe (ePe1)2 NPe . in Eq.1 e Therefore.13.25) (4.13.25. 4.24) It is also possible to go through the similar procedures with a central differencing form of the convection term in Eq. 24 j.13.22 we can obtain the effective Peclet number Pe of the finite difference solution in terms of the real Peclet number P as follows. 4.25) . and the effective diffusion constant for the numerical solution will be obtained.1 (4.22) Comparing Eq. 4. 4. 4. Eq. 4. 2 ii P ( (1+P) (1+) N _ 1 N 4).+1 i + ) il 0 (4. Pe = in (1+P) (l+P) iIp2 N (1+P) . The in analytical solution Eq. 4. + 24 S i i+1 i1 =e e l NP  1 (e P P 1) 2) ) ( (e 0 (421) The finite difference solution Eq. i+l 2Ax il S S2ax 4) i+l 24i + i 2 i (4.12 is substituted for 4i Eq.20 is also substituted for 4.56 solutions will be reinserted to the finite difference equation. 4.23) (4.
4. Pe = kn (2P) 2+P (4. uLx D De Pe  (4. The convection terms in . 4. As in the previous section the velocity u is assumed to be positive and constant. it helps understanding the origin of the truncation error diffusion.26) where 2 < P < 2 The effective diffusion constant De can be readily obtained from the following relation.57 The effective Peclet number for the finite difference solution of Eq.2 Another Approach For a One Dimensional Truncation Error Diffusion The truncation error diffusion in a one dimensional problem can be evaluated in a different approach. Eq.26 may be a good indication of the truncation error diffusion occurring in a general one dimensional problem.27) In general.25 can be derived as follows. the transient and source terms and the mixed type of boundary conditions will not affect the numerical diffusion appreciably if they are not dominant terms in determining the profile of 4.24 and Eq. Therefore. 4.1. Although this approach is approximate and overpredicts the truncation error diffusion. 4.
30) Both Eq. i.26. the diffusion is dominant over the convection. Pe also goes to zero resulting in no diffusion error in Eq.29) i+ The truncation error diffusion can be quantified by subtracting Eq. 4..5 may be approximated as follows.28.29 from Eq. 4. 4.28 and Eq.e. 4. i. the truncation error diffusion constant for a convection dominant problem may be given approximately as follows.31) ND .28) Ax (4. D uAX 2 (4. the convection is dominant over the diffusion. It can be shown that as P goes to zero. 4. the profile of 4 will be linear and the truncation error will be reduced to zero. 4. If the Peclet number is small. u i Ax i1 u i+ a4 u() ax u(a) ax (4. ae u( x where 0 <f < ae + u() ax i+  u~x a2 ) .e.29 hold only when the Peclet number is large.58 Eq. 4.( i+f 2 (4.4 and Eq. 4.24 and Eq. Therefore.
3 Truncation Error Diffusion in a Two Dimensional Problem There are two sources of numerical diffusion in a two dimensional problem.33) . A steadystate. It will be shown that the truncation error diffusion in the direction normal to the flow cancels out and there is only a flow direction component left. u i 6x ilj + v i by ij_a(i+1 2 4x i + il1 +ij+l 2i 2 + 0iji by (4. The formulas for the truncation error and crossflow diffusion constants derived in this chapter reveal that they are approximatly of the same order of magnitude. 2 u + v 2 + +( ) (4.32 is finitely differenced using donor cell scheme for the convection term.1. two dimensional.59 4. one is the truncation error diffusion and the other is the crossflow diffusion. conservation equation with no source term is given in the following. Since the gradient of 4 in the flow direction is negligible in a convection dominant problem. the total diffusion quantity of 4 due to the truncation error is not significant in comparison with that due to the crossflow diffusion.32) = x 3y Equation 4.
33.v( ay ) ij .35) . transformed to the coordinate ( . 4. ij Ax 1 1x Axhy 'i1 i Ay iji 1 +AxAy fj+ j i+ i axx dt + Yj+ yj+ yj xi+4 xi+ a (v)dx dy ay Xiih + u(t) x i+ a =U(ax) )3x .y) is by rotation.n) ( = x cos6 + y sine I = x sine + y cose (4.60 The approach in section 4.2 is applied to Eq.j+ 2 2 ay 2 a2 2 A Ax U(a 2 ) + bay U( 924 )+ Ax 82 2 axay 2 ax2_I = AX Ax A 2 axay va a yyaxa 2 ay 2 S[uAxax + (uAy+vbX) +v = .1.x[cos + (cose tane 1 + sine) + sine tane ] ax xY (4.j+ + v(a) ly i+h. then the truncation error of the convection term is given as follows.34) where tane = Ay/Ax tane1 = v/u The coordinate (x.
37 shows that the diffusion component in the n direction (normal to the flow direction) is zero and the cross differential term also vanishes when e is equal to e1 Therefore. dominant truncation error diffusion occurs When only in the flow direction. 2 (A)  [(sine tane1 + cose) 2 + a2 (sine + cosetan6 1 ) a + (0) 4 an U" (4. e is equal to 0 or 7/2. 4. 4. the problem is basically one dimensional and the truncation .32 is reduced to the following form. 2 2 2 + a = ( 2 ) (4.61 where 5 is the coordinate in the flow direction. y v e Sx Then Eq.34 will be transformed to the coordinate system (t.37) Equation 4.n) given by Eq. 4.35.36) where U = /u2v 2 Now Eq.
4. [67. In this section the origin of the crossflow diffusion will be illustrated and the corresponding diffusion constant will be quantified so that they can be used in the corrective scheme of section 4. therefore it exists only in two or three dimensional problems when the flow direction is not aligned with the mesh configuration.2 Crossflow Diffusion Most of the confusion about the nature of numerical diffusion comes from the fact that there exists an additional source of false diffusion. that will be named here as the crossflow diffusion. The hot fluid will come out of the top surface and the cold fluid out of the right surface.l1(A).1. 4. 4.62 error diffusion occurs in the flow direction as in a one dimensional problem. the hot and cold fluid enters the mesh from the left and bottom surfaces at 45* angle.3.1(A) is transformed into . In the donor cell scheme Fig. The origin of the crossflow diffusion is illustrated in Fig. which shows a single mesh with pure convection. In Fig.4.68) and it was clarified that this is the dominant source of the error in most multidimensional problems. The crossflow diffusion comes from the multidimensionality of the problem. tional false diffusion was explained by Patankar [52] and Stubley et al. 4. It is entirely different Recently this addi from the truncation error diffusion.
63 Hot Mledium Hot 7 Hot Cold (A) Cold (B) Cold (C) Fig. 4.1. Illustration of the crossflow diffusion in a single mesh with pure convection .
. Another illustration of the crossflow diffusion is given in Fig. 4. and the sum of the values This distribution along the points A1 . diffusion occurs in multidimensional donor cell differencing of the convection term. at those points is exactly equal to 4. CAB which looks like a long onedimensional rod. A 3 . 4. In the donor cell scheme it is distributed all . A 2 . . an appreciable amount of false program user.and intermediate temperature fluid will come out of the top and right surfaces. in The numerical results Fig. .1(C) by a Therefore. 4. 4.2 which shows a row of meshes aligned in the xdirection.1(B) where the velocity components normal to the surface are considered. The value of 0 at the point A should reappear at the point X because there is no physical diffusion. 2 (lp)[l+p+p + .1(B) are again interpreted as Fig.3. Then homogeneous mixing occurs in the mesh. . of the value of 4 causes the crossflow diffusion. 4.64 Fig. ] = (lp)4 = where u .&x u v ax Ay The effective diffusion constant for the crossflow diffusion will be calculated in the two dimensional Cartesian coordinate in Fig. This is a pure convection problem Consider the fluid element The fluid with no physical diffusion.
Illustration of the crossflow diffusion in a row of meshes aligned in the xdirection with pure convection .65 (1p)4' x p (1p) 2 p (lp) p3 (lp)4 p4 (1p) A5 Ay A Ax 2 4 y p t* U where p = Ax Ay Fig. 4.2.
Geometry for explanation of the crossflow diffusion in a two dimensional Cartesian coordinate . AY B Ax t^ t where t t0 = tane = v/u u/Ax v P U = 3/U tane= Ly/Ax Ax by Fig. 4. V A2 e A.3.66 4.
4. p4A and (lp)4A and appears at the points Al and A 2 . Gradient) A /k 1 =A of 4 Current 2 2U of 4 Therefore. k2 = Ax sine/sin(el+8) k3 = Ax sine 1 /sin(e 1 +8) After substituting appropriate expression for each term. and is translated The value 4A at the point A should reappear at the Point A'. sine cose cos= sx Uacos61 sin(O+e ) I DRHSS . The diffusion occurs both to the The diffusion to the right hand side is considered first in the following. right and left hand sides. Diffusion A t1 .67 element was on the line CAB at time t in the velocity direction. the following right hand side diffusion constant results.t 0 PA £3 (Current) 2 1 constant ) where 1 = (Gradient) = U3 Lx/cose. v. 4. The travel of the fluid element is analyzed in Fig. which shows that the value 4A diffuses out along the fluid element. In the donor cell scheme it is divided into two portions.
Ya.4. 4.. Geometry for explanation of the crossflow diffusion in a rotated two dimensional Cartesian coordinate .68 ti .ct t2 t3 Fig.
The crossflow diffusion constants are the ratios of the current and gradient of 4 in the x and y directions.38 is the effective diffusion constant when the diffusion is confined to the one dimensional fluid element. 82 x 9x 2 2 By 2 2 The current on the one dimensional fluid element can be factored into the x and y components. sine cose RHS LHS cos1 sin(+8) (4. D D = uAx(lp) = VAy p (4.38) Equation 4.39 is identical to the following simple expressions. sine cos6 D = UAx sin cos8 x sin(e+e 1 ) 1osel 2 sine cose sin 81 D = UAx y sin (8+8 ) cos1 1 (4.39) Equation 4. There are two diffusion components in a two dimensional problem as follows.69 Repeating the same procedure for the left hand side we can find that the right and left hand side diffusion constants are equal to the following.40) .
In the donor cell p' y scheme the value 4 is divided into three portions px 0 . 3 0 when tan 6 = tan6 1 . that is to say.6 the value at the point 0 is 4 and that value should reappear at the point X where the velocity vector meets the plane ABC. 4.40 and Eq.42 give the same result.42) Equation 4.6. 4. < 6 < 81 or < . In Fig.41 shows that DRH S (or DLH S ) is maximum at an Tr7LHS angle of 61 < 6 e between 4 and e1. D DDM = UAlxy sin2e= 3 3 4(by sin36 + Ax cos36) UAx tan6 1 sin26 3 4(tan8 sin6 3 + cos3) (4.38 should be differentiated with respect to 6 in order to find out the velocity direction with maximum crossflow diffusion. 4. When both 6 and 61 are equal to . Eq.70 where p = u Ax u v By Equation 4. RHS D = It can be shown that. 4. 4.42 underestimates the crossflow diffusion as shown in Fig. Now the analysis will be extended to a three dimensional case in Fig.42. 4. De Vahl Davis and Mallinson [24] also derived the effective crossflow diffusion constant and their result is given in Eq.5. (4.41) Equation 4.
0t Dcf UAx de Vahl Davis & Mallinson Huh 81=8C 1. Comparison of the prediction formulas for crossflow diffusion constant by De Vahl Davis and Mallinson and Huh . 0.0 0 0 .71 Cross flow Diffusion 2.5 80 \ % % 4 1.0 150 300 450 600 750 900 Fig.5 I I 0. 4.5.
p Fig. 4.72 z C V Az X 0 o a X AY Ay B y x z C p .6. Geometry for explanation of the crossflow diffusion in a three dimensional Cartesian coordinate .
43) Since the point X is on the plane ABC.and nA unit vectors along those directions. XC. This is the socalled crossflow diffusion phenoThe currents due to the crossflow diffusion are . CX which are denoted as £i' z2' k3 can be obtained as follows. S= 4 + y z zAx 1 cos+ + cos Ay Ax + cosy Az (4. £4 cosy) nB .73 z occurring at the points A. XB.44) 4 Therefore. £1 = AX =/(x4cosa)2 + (4cos)2 4cos) 2 = Ax 2 + 42 2Ax£ 4 cosa . B. the following relation holds. along the directions XA. BX. n C denote The coordinate of the Any point on the plane ABC should satisfy the following Eq. 4.45) The lengths of AX. point X is given as follows. cosa cosB cosy (4. and C instead of at the point X.43. X(£ 4cosa. t4 cos8. S+ Ly Ax + Az = 1 plane ABC (4. menon.
n C can be obtained in terms of their components in x. Therefore.46) + The unit vectors nA . y. nB' n C can simply be given as the products of the transported quantities and velocities. 2 o C 14cosy] = 1  £ 4 cosa.48) The currents in the directions nA . Z 4 cos8. nB. z directions.OX (4. (Ay4 cosS).47) etc. (Az 4 cosy)] (4.74 k2 = BX = /( 4cos)) 2 + (Ay_4cos) 2 + (k 4 cosy)2 =  y2 + + 42 L 4 £3 = CX = /( 4cosc) 2 + (£4cos8)2 '' + (AzL 4 cosy) / z + £42 2Azi + 4 cosy (4. I (Current)xAi XAI = px Px C1 1 Lt . 4 cosy] 1 nB = [i4cosa.  n A = XA +1A  OA .
(4 pz x co s a ) ct 4 + at ( 4 cosa) 4 .50) Now the x direction crossflow diffusion constant may be given as the ratio of the current to the gradient of 4 in the x direction./ Ay 4/6x Q/by (Grad)z = 4/Azl (4.75 I(Current) XB = Py p2 t I(Current) XCI where At = k /U z '3 At (4./ Ax (Grad) = 4. (4.49) The gradients of 4 in the x.51) p Ax D x = lt (Axk 4 cosa) + p Ax . (Grad)x = 4. z directions are given as follows. Px D (Ax£ + Py D + nC(4 (Current)x 4 cosa) c osa) At ( 4 (cosa) (Grad)x = 4/Ax Therefore. y.
1 = uAx(lpx) (4. although none of them has been successful enough to be accepted widely.76 p D x Ax At A2 At t cosa(p +p +p) 4 x y z SxA At Ax L cosa at 4 (4. . This section is a review of the schemes that The have been used to eliminate the numerical diffusion. etc. finite element method. tensor viscosity method. Dy = vLy(lpy) Dz = wAz(lpz ) The rotation of (4.54) 4. we can simplify the result as follows.53) The crossflow diffusion constants in the y and z directions can be obtained in the same way.3 Discussion of Skew Differencing and Corrective Schemes The origins of the numerical diffusion and their effec tive diffusion constants have been given in the previous sections. p Ax D = ux[ £4cos x . most important ones may be the skew differencing. indices will also give the same result.52) Inserting the expression for At given by At = £4 /U.
Chang's method 1161 and Raithby's scheme 1533 are of a skew differencing type while the tensor viscosity method [25] and Huh's corrective scheme that will be introduced in section 4. In the skew differencing scheme the 5point relationship becomes a 9point relationship.II1~LI1I~~ ICII1_1___~1___*_____11__ 77 Most of the schemes in Table 4. but directly along the velocity direction. However. In the skew differencing type schemes finite differencing of the convection term is done not in terms of the x and y directions..3.1 can be categorized into two types. The crossflow diffusion is not eliminated entirely in the skew differencing scheme.~_ 1 i_..^ILli. The donor cell treatment of the convection term results in a 5point relationship in a two dimensional problem. The corrective scheme is to reduce the diffusion constants in order to compensate for the additional false . S. the interpolation between neighboring points may still give some crossflow diffusion. the interpolation should be done between neighboring points to obtain the value at the upstream point. Basically the inclusion of four additional corner points contributes tomore accurate numerical modelling of the convection. but just appreciably lower than that of the donor cell scheme.4 fall in the category of a corrective scheme.. Since the upstream point in the velocity direc tion does not necessarily fall on the mesh point where the quantity under consideration is defined. skew differencing and corrective schemes...
£Ebzhau t aL/PATANKAR Lilinpion Orlandi Priddin Wada cit l.ordc.Highe cdtr upwi:nd :"'ie Cdi't..j. Esposito Glas and Rodi Huffenus Lnd Khali.hod Lillinron Ozlandi Snd hod m u macsh? Yes Yes Yes Pc = 10' only Yes Yes Yes Pe = 10' only es mesh 23 x 11 40 x 40 65 41 21 41 x x x x 33 21 11 21 di 'frencc Gt:Iz:kin .1QUICK t ETovh ci .inte Conribo:r/me.ad.zky Chen Rue) ct &L Schonaucr Sykes Wad ct aL 41 X 21 21 x 1 40 x 40 100 x 50 81 x41 100 x 50 41 4l 3E 80 x x x x 21 21 19 40 .aptjve me. upwind f~iite di Teretn atl.hod Mesh *txLnsformation rr./LLUE Huint/QUICK Wilkes/LUE Lillimpon/SUD LiiL. viscosity method Sel. List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research.1.78 Chss' iction Cen:ered . 1981 [65) .ronLUD Lillinion/VUDCC Nuana.thod of chaaceriics Fiite anuryic method Tenso. Grandotto Hickmott Firs.thod LLxW'en doff  21 x11 80 x 400 Table 4.tnce Vuer:o: upwind Yes Yes Yes Yes 2) x II 21 x ]1 21 x 11 21 x 11 41 x 21 41 x 21 21 x 11 81 x 41 28 x 14 Se Tzb)t 2 Up'nd ft(e elemmnt !.e elements Clifft Dontcl t al. Wilkes Elbahu ct l.
and D z in Eq.1 Raithby's Scheme Raithby's scheme [53] is an Eulerian type of approach. D..54. 4. 4 = C1 + C2 n = C1 + C2 (  (4. although the validity of its correction formula is not clear. It is therefore a conservative scheme. Since the crossflow diffusion is dominant over the truncation error diffusion and the effective diffusion constants for the crossflow diffusion can be predicted theoretically.55) where n is the distance normal to the flow and U = /u 2 +v 2 .54. The tensor viscosity method is one example of the corrective scheme.53 and Eq. 4. 4. The finite difference equation for Raithby's scheme is derived on the assumption of a linear profile of 0 normal to the flow and a uniform profile in the flow direction. The correction formula The diffusion con used here is Eq.53 and Eq. 4. 4. stants in x. the corrective scheme can give a numerical solution which is almost free from numerical diffusion. where the balance equation is set up for a given control volume by considering convection and diffusion at the control surface.3. y and z directions will be reduced by the amounts of D .79 diffusion. The corrective scheme is valid only if the effective numerical diffusion constant can be predicted accurately.
the calculation domain showing grid lines. 4.7. A segment of. control volume and notations in Raithby's scheme .80 Fig.
'ij US (4.81 The convection quantity at the control surface is calculated from the upstream value in the flow direction and the upstream value is obtained by interpolating two neighboring points. This should be repeated for the four control surfaces of every mesh in a two dimensional problem.2 Skew Differencing By Huh Skew differencing scheme here treats the convection term in a Lagrangian way.3. The convection term is differ enced in the velocity direction between mesh points.= U ax dy SC (4.56) where U = /u is the coordinate in the velocity direction and 2 +v 2 . it is much simpler than Raithby's scheme and becomes conservative in a unidirectional flow. 4.57) where the upstream value €US may be calculated by . Although this scheme is nonconservative. The inclusion of the four additional points also complicates the structure of the coefficient matrix in an implicit scheme. not in terms of the convection that occurs at mesh interfaces. u 4 + 'v. which may be too complicated and time consuming for a practical purpose.
Some of the results of this skew differencing scheme is given in section 5. uv (4.58) t2U= Eq. 4. D are derived . Dx = uAx(lpx) D Yy = Ay (lpy ) The expressions for .82 interpolating two neighboring points. D . The validity of this scheme is based on the fact that the crossflow diffusion can be predicted accurately. DD z in x. the tensor viscosity [25) T = Lt uu u2 The expression for is given in the following. 4. 4.2. DD y .54.3. DDx. z directions to compenx sate for the crossflow diffusion effect. 4. the crossflow diffusion constants D in section 4.53 and Eq.2. y.3 Tensor Viscosity Method The tensor viscosity method treats the numerical dif fusion constant as a tensor quantity.3. 4.58 for the numerical diffusion constant is apparently not consistent with the results in Eq.4 Corrective Scheme By Huh The corrective scheme is to use the reduced diffusion constants.3.
Dz at the center of a mesh where all quantities are defined except the flow field. .1 Mesh Point Implementation The mesh point implementation is to assign the crossflow diffusion constants Dx . In a coarse mesh or recirculating flow field the latter implementation strategy is definitely preferred.3. Dy . Two implementation strategies have been tested so far. the mesh point implementation and mesh interface implementation.These two become identical when the flow field is uniform or a very fine mesh is used so that the flow change over neighboring meshes is negligible.+ . After the crossflow diffusion constants are assigned to every mesh point. The velocities at two boundary faces are averaged to get the representative velocity.59) v u . 4.4. which is used with the cell dimensions to get the crossflow diffusion constants. they should be averaged again between neighboring mesh points to calculate the crossflow diffusion current at mesh interfaces.+ w Ay Lz x etc.._IIU__lpULI~~I___~lLlt i~Lli ~i ~iL~pn~sl. 83 D z = wAz(lp z ) where px = u (4.
so that they are proportional to V1 and V2.V2) are used to give (DxlDyl) and(Dx 2 'Dy2 ).4.8. There are three possible cases that should be treated independently in a two dimensional case. The velocity U is split into two components. The third case is expected to have a flow split as shown in the figure. The right interface will have the crossflow diffusion constant (Dx1+Dx2 ) and the top and bottom interfaces. . The two velocity sets (Ul.84 4. U and V.2 Mesh Interface Implementation The mesh interface implementation is to assign the crossflow diffusion constants directly to the mesh interfaces where the crossflow diffusion current should be calculated.3.Vl) and (U2. The same logic can easily Appendix C includes be extended to a three dimensional case. can be directly used with cell dimensions in Eq. Dyl and Dy 2 . and the resulting D x and D y will be assigned to each interface.59. For the first two cases the cross flow diffusion constants are zero because the velocity vector is aligned with the mesh configuration. as shown in Fig. U1 and U2. 4. For the second case the two outward velocities. The mesh interface implementation is always recommended although it has a more complex program logic than the mesh point implementation. Note that the crossflow diffusion constant for the interface with an inward velocity does not have to be considered in that mesh because it will be considered in the neighboring mesh. 4.
4. Three possible flow configurations for mesh interface implementation of the corrective scheme in a two dimensional problem ._~I~~ _~~~~_I__~__~~ __ 85 (1) (2) (3) Ul U2 V2 VI U U V14V2 VI+V2 V2 U2 = UV2 U = Ul + U2 TV2 Fig.8.
60) Explicit (lD): n+l I n i At + u (i i x n i. 4.4 Comparison of Explicit. Implicit and ADP schemes There are three typical solution schemes. Eq. which is a parabolic partial differential equation.86 the fortran program for calculating crossflow diffusion constants in an arbitrary flow field. 4. Explicit.61) Implicit (lD): n+1 i n 6t At i u ( Ax n+l i  n+l i1 ) = D A 2 n+l i+2 i+l n+l + i n+l i i1 ) (4. The pros and cons of the three schemes will be compared in the following four viewpoints. 4.60. Implicit and ADI (Alternate Direction Implicit). conservation physical constraint accuracy stability 2 ax 2 2 y 2 +ux +at + 7 + T7) + S sz (4.) = i1 n D Ax ( i+1 2i l n n + n ) + iS i1 (4. for the general conservation equation.62) . 1. 2. 3.
+ 4). when the net inflow or outflow at .87 ADI (2D): n+ ij (A) _n ij In+ + u ij _ cn+ in + v n ij1 At 2 Ax Ay )n+ i+lj 24 n+ Ax 2 +n+ n l + D j+l n ij 2 Ay n+l ij n iji + S n+l ij n+ ij n+ ) ij n+ i1 n+l ij1 At 2 Ax Ay n+ =D i+i S2.4. 2 Ax n+l 2 * ij v n+l ) ij n+l Ay 13j1 n+1 ij+l n+l 1j1 At Ay2 + S (4. n+ n+ + 4n+ n+l + D ij+l  2 n+1 + 4n+l i i1 + Ax2 ADI (2D) : * Ay 2 (B) n *i * 1 j + 1 1 i+l  At _ S +Ax 24..63) 4. or affected only by the source term contribution (SO).1 Conservation The conservation principle is that the total amount of 4 should be conserved (S=0).
88 the domain boundary is equal to zero. When we are concerned with a steady state solution. Although they usually give . All the explicit. there are some nonconservative solution schemes like ICE (Implicit Continuous Eulerian). Although a conservative form is preferred to a nonconservative one. implicit and ADI schemes can be made conservative by using appropriate differencing forms for the convection term. A scheme may be both stable and nonconservative and also both unstable and conservative because the conservative property is independent of the stability.4. + CI 4. where the convection terms of momentum equations are linearized in a nonconservative way.2 Physical Constraint The physical constraint is closely related with the stability in a practical sense. the conservative property may not be crucial. the conservative property is as important as the stability because the error from nonconservation may accumulate over many time steps. the same exchange term should be used for neighboring meshes. In order to guarantee conservation. For a long time transient. however.
The Courant condition and Hirt's stability condition [36] that the finite domain of influence should at least include the continuum domain of influence are good examples of the physical constraint condition. The laws of physics impose some restrictions on the numerical scheme to get physically reasonable solutions. )C  4R Ax 6y 4R C Ax Exchanged quantity of 4) by diffusion for time At = DR Ax CAy At After the exchange of the diffusion current. new values of 4 are given as follows.89 similar constraints on the time step size and mesh spacing. they are from entirely different origins. R C AxAy = CAxy + D R Ax Ax CAy C y At At ¢4AxAy = 1RAxAy .D .2.1 Diffusion There is a maximum net flow that can occur between neighboring meshes by diffusion. 4.4.
scheme given in Eq. t' > Q' .64 or 4. .64) Eq. 4.90 Since the relationship.63 is unconditionally stable. R AxAy(4) C) > 2D (R Ax CyAt 2 DAt X2 < 1  (4. d 2D t < 1 (4.65 may be a too conservative criterion in a real problem. we have the following physical constraint. When the physical constraint is ignored and the time step size greater than that given by Eq. there will be a damping oscillation in the solution when the time step size is greater than that given by Eq.65 is used.64 may be extended to a ddimensional case as follows.65. should still hold. 4. Therefore the physical constraint should be respected in addition to the stability condition to get a meaningful transient solution. however. the solution will be unstable or stable with damping oscillations. Both of the oscillation and instaThe ADI bility should be avoided in a transient problem.65) Ax Equation 4. 4. 4.64 or 4.
91 4. At < 'x u (4.66) When a time step size greater than the Courant limit is used. the damping oscillation or instability may occur as in the case of diffusion. 4.B 4y S The explicit and donor cell treatment of the convection term gives the convection quantity at interface S as ulAAtAy. It is based on the assumption that 4A is the repreIf At is greater sentative value of the control volume A.3 Accuracy Realistic interpretation of a finite difference equa tion comes from an integral form of the conservation . Then uAttay is no more an appropriate expression for the convection quantity during At. Therefore. a portion of the control volume C will cross the interface S. than the Courant limit.2 Convection C A u B C A ).2.4.4.
9. ~~CIIL Is CIIII r r I I I I I! US Ds > I r I US DS Donor Cell Central Differencing Reverse of Donor Cell where US(Upstream) DS (Downstream) Fig. Profile assumptions of Q in space and time for various differencing schemes .92 n+l n+l 4' I I. 4. I I t i r _ I I [ n n+1 n n+l _ 0 It Explicit Implicit CrankNicholson $i+.
whether the error introduced at a certain step will increase or decrease as the calculation goes over to the following steps. The infinitesimal approximation of differential terms is not appropriate in most problems of our consideration. In other words. The stability can be checked by the theorems about matrix eigenvalues or Von Neumann analysis.4. Since the exact profile depends on the Peclet and Strouhal numbers of the governing equation.93 equation. 0(Ax 2 ) is The also not appropriate for finitely large At and Ax. appropriate schemes should be chosen according to those numbers or weighting should be done between neighboring time steps and mesh points to increase the accuracy of the solution. the accuracy depends on the profiles 4 in time and space because the exact profiles assumed for are not known a priori. If its absolute magnitude is less than one. accuracy depends on how well we can compute the convection and diffusion quantities at control surfaces over the time At.4 Stability The stability is a mathematical problem. the error propagation is suppressed and the largest error contribution will be from the previous time step. the accuracy is usually the least important among the given four considerations. It depends on the maximum eigenvalue of the iteration matrix. 4. The accuracy expressed in terms of O(At). The Von Neumann . However.
2 2i + D i+1 i n 2 i i1 (4. implicit and ADI schemes using the Von Neumann analysis.94 analysis is to expand the solution of a finite difference equation in Fourier series and check the decay of each mode separately. 4. 1 4n = E~nei where I = . lowing sections.67) The solution 4.4.4. but its algebra may get complicated. (4.4.. is expanded in Fourier series as follows.1 Explicit Scheme The Von Neumann stability analysis of a finite difference equation is simple in its idea.68) . stability criteria will be derived for the explicit. The stability is not a sufficient conIn the fol dition for an acceptable numerical scheme. n+l i n €i+i I + u i+1 n ii 2Ax i_1 lt n n i+l .1.1 One Dimensional Central Differencing of Convection Term The finite difference equation is given for a one dimensional general conservation equation with central differencing of the convection term. 4.4.
the function . 4. f(t) = i 12 = (4d2_ 2 )t2 + (Sd 2+4d )t + C 2+4d 24d (4. 2 11 = [1 .67. [i .2dx(cosex) 2 + [Cxsinex 2 < 1 (4. in Eq.68 is substituted results.69) Dbt Dt 2 Tx.71) The function f(t) will be defined as follows. 4.67. Therefore. the following expression for = lCxsin xI where cx d x uAt A tx  2d (lcos x) (4. mode to achieve stability.72) In order to get the stability of Eq.70 may be rewritten as follows by defining cos6 x as t.70) Equation 4. 4.95 The absolute value of should be less than one for every When Eq.2dx(t) where t = cose 1 < t < 1 ] 2 +C 2(1t 2) 1 (4.
f(1) = 0 (4. 4. Therefore.1] Equation 4. (1) 4d x 2 Cx 2 > 0 x : concave upward : linear 2 (2) 4d 2 C 2 = 0 (3) 4d x 2 Cx 2 < 0 : concave downward.73) for any t in [1.74) One of the two meeting points with the taxis is the point t=l because f(l) is equal to zero. The parabola f(t) always meets the taxis because the discriminant is always greater than or equal to zero.1].75) The parabola f(t) may have three distinct shapes according to the sign of the coefficient of the second order term. f(t) < 0 (4.73 can be solved by a graphical method using the characteristics of a quadratic equation.96 f(t) should be less than or equal to zero for any t in [1.10(1). In case (1) f(t) is a parabola concave upward and f(l) should be less than or equal to zero as shown in Fig.2d )2 > 0 (4. . D(Discriminant) = (C 2 .
according to the sign of the coefficient of second order term .3 0 0 /1 i2 2 4d 2d x x 22 4d c xx t / (1) 4d 2 c 2 >0 x x (2) 4d 2 c =0 x x 2 2 (3) 2 2 4d2c2<0 x x Fig.. 4. r I I t i t 1 1 t 0.10. Three distinct shapes of the parabola. f(t)..97 f(t) f (t) f (t) 'I I 4.
2 One Dimensional Donor Cell Differencing of Convection Term The following Eq. we get the following stability conditions.10(3).76 and Eq.76) In case (2) f(t) is a straight line and f(l) again should be less than or equal to zero. 4. 4. 4.11. 4.4. .77.dx) is shown in Fig.2d Axis of symmetry: t = 4dx dx > C 2 x > . 4d 2 C 2 4d C > 0 < 0 : : 0 < d dx > < 2 Cx The stable region in the coordinate system (Cx.C (4. 4.98 f(l) = 8dx(2dx1) < 0 0 < d < (4.78 is a finite difference equation with donor cell differencing of the convection term. In case (3) the parabola f(t) is concave downward and the axis of symmetry should be on the right hand side of the point. t=l as shown in Fig.4.1. 4d 2 . 4.77) Summing up the results in Eq.
11.99 d x 1 1 0 1 dx = C2 x 2x C Fig. 4.d ) for an explicit scheme with central differencing of convection term in a one dimensional problem . Stability ondition in the plane (Cx.
100
n+l
i
n At
At i
n
+ u i Ax
n
ii
n
= D
i+l
n
i
n
ii
Ax
2
2
(4.78)
where the velocity u is positive. The Von Neumann analysis of Eq. 4.78 gives the following stability condition.
2
It
1(C x +2d [
x
)(1cose x ) 2 + [C sine ) x
1
(4.79) (479)
can be shown that the value of IC12 in Eq. 4.79 does not
change when the velocity u is less than zero, if the Courant number, Cx, is defined such that it is always positive. The function f(t) is defined as follows to make it easier to solve Eq. 4.79.
f(t)
=
12
1
S[(C x+2d)
2Cx
]2 t
+ [2(C x+2d )2 2(C +2d
) ]
+ 2(C x+2d )]t (4.80)
+ [(C +2d ) +C
The function f(t) in Eq. 4.80 should be less that or equal to zero for any t in [1,1] for the stability of Eq. 4.78. f(t) < 0 for any t in [1,1] (4.81)
The parabola f(t) meets the taxis at the point, t=l. f(1) = .0 (4.82)
101
Equation 4.81 canbe solved by the graphical method as in the previous section. 2 2
(C +2d )
 C
> 0
:
0 < Cx +2d X < 1 
(4.83)
(c x +2d x)
2
C
2
x
< 0
d
>(C 2C ) dx x
(4.84)
The results, Eq. 4.83 and Eq. 4.84, are shown graphically in Fig. 4.12.
4.4.4.1.3
Two and Three Dimensional Donor Cell Differencing of Convection Term
The two dimensional extension of Eq. 4.78 is given in the following. n+l ij At n D . i+D 3 24
Lx
n ii
n + u ij Ax n
n i1 n
n + vii by
n iji n
+ 4
A2
1il3
n n 4  24 + D2+l ,_2
ay
+
ii
(4.85)
The Von Neumann analysis is applied to the finite difference equation, Eq. 4.85, yielding the following result.
 1 = C x (le
lex)

C (le I
y
(lcos8 y) _ 2d (csex ) x x
(4.86)
 2dy (1cosey)
102
d
x
1 2
x
1 2xx
1
d
1 2 1 C Cx)
Fig. 4.12.
Stability condition in the plane (Cx,d x ) for an explicit scheme with donor cell differencing of convection term in a one dimensional problem
103
where
C
x

Ax
C
 vat
d

DAt2 x2
x
Y
DAt Ay2
The Courant numbers, Cx and C y , are defined so that they are always positive, independent of the directions of the velocities u and v. Equation 4.86 can be simplified
by defining al and a 2 as given in the following Eq. 4.87. = 1 + al1 where a= 1 C x  2d x + (Cx+dx)e x I6
x +
a2
(4.87)
+ de x
le
x
S
=
2
C
Y
2d
+ (C +d )e YY Y
I6
x
+ de Y
Iy
The stability condition is that the absolute magnitude of 5 should be less than or equal to one for any values of 8x
e and By.
From Fig. 4.13 and Fig. 4.14 it can be seen that
indea1 and a2 should satisfy the following conditions pendently. a + 1
4
(4.88)
(4.89)
 .13.1 SRe 2 al +2 1+a +a 2 Fig.104 Im \ . 4. The region in the complex plane where (a +c ) should exist for the stability of an explicit conscheme with donor cell differencing of vection term in a general two dimensional problem .
. The region in the complex plane where the amplificatior factor. . should exist for the stability of an explicit scheme with donor convection term in cell differencing of a general two dimensional problem . 4.14.105 . Fig..Re Both al and a 2 should be in the shaded circle to ensure that Ial+a 2 +1I< i.
15.91) . At. 4. the stable time step size for a given mesh spacing. 4. which depend on both the mesh spacing. When Ax and At are already given. 4.88 and 4. I 1 2 + (1+)e Rx Rx 2 + (1+)e a1 = Cx [1 + Ie 1 3 .4. The stable region shrinks on a linear scale with dimensionality.4. A three dimensional case can also be solved through the same procedures and the results are given in Fig.88 and Eq. the stability However.dx) and (C y.e Y) (4. which is independent of the time step size.16. 4. Cx and d . the stable range of Cx should be determined in terms of the cell Reynolds number Cx/dx'. The results are given in Fig.d) should be in the shaded region for stability.89 and al and a2 may be rewritten as follows. our usual concern is can be checked directly. Therefore.90) a 2 = C [1 y R R Y + R (4.106 Both Eq. where the points (Cx.1. At. and time step. The stability condition is again given by Eq. Ax. 4.e Rx .89 can be solved by the graphical method as in the previous sections.4 Stability Condition in Terms of Cell Reynolds Number The stability conditions in the previous three sections are in terms of the dimensionless numbers. 4.
dy) fcr an explicit scheme with donor cell differencing of convection term in a general two dimensional problem.15.idition in the planes (Cx. .107 d (d ) d = x 2 1 x + 1 4 C x(C ) 16 d= C2_ C x 2x x Fig. 4.dx) and (Cy. Stability cc.
for an explicit scheme with donor cell differconvection term in a general three encing of dimensional problem . Stability condition in the planes (Cx.108 d xy(d or d z ) Cx d dx + 1 1 Cx (C d! S3 2 . 4.dx) etc.16.(C Cx x y or Cz) z d ) Fig.
) RX (1cose )) x + [C sine] x x < (4.95.94 ) D(Discriminant) = C 2 (K2C ) 2 > 0 (4. f(t) = C (1+K) 2 (lt) . (1) Cx 2 K2 1) > 0 concave upward .92 can be simplified as follows. 4.95 ) The function f(t) satisfies Eq.R and x t = cosex The stability condition is now that f(t) should be less than or equal to zero for any t in [1.92) Equation 4.94 and Eq.1] f(l) = 0 (4.C K(lt) + C ( l_1t = (C 22 KC 2 2 )t2 + (2C 2 22 22 K +C K)t + C 22 K  C K + C 2 < 0 (4.C (+2x . 4.90.__LI___IIII__LI*X__ ID I 109 The following condition for Cx and Rx can be derived from Eq. [ . 4.88 and Eq. 4.93) ) where 2 K = 1 +. and has distinct shapes for the Following three cases.
17.110 (2) C 2(K 21) = 0 linear (3) 22 C 2(K 2_1) < 0 concave downward For case (1) and (2).97 are shown in Fig. f(l) should be less than or equal For case (3) the axis of symmetry of the parabola The should be on the right hand side of the point.96) R 1 x < 0 < C x < (1+2) Rx . 4. solution procedure is similar to those in the previous sections and the results are given in the following as. to zero. therefore the subscript x may be replaced with the subscript y in Fig. 4. The stability condition in Fig. 4. (4.17.97) The results in Eq. The same result will be obtained for the ydirection.96 and Eq. V f .17 is applicable to all general 2D problems.98) Ax . t=l.yu < 1 (4. R > 0 x : 0 < C < x 2 2 2(1+) x and (4. 4. although it can be relaxed if there is a constraint on the velocity direction and mesh spacings such that the following' relation is satisfied. 4.
17.Cy) for an explicit scheme with donor cell ditferencing of convection term in a general two dimensional problem . 4.111 x (C y C x  2 1 2 (1+ L) R x C = 2(1+ ) x 2 0 R (R x y Fig. Stability condition in the plane (Rx.Cx) and (Ry.
Cx) for an explicit scheme with donor cell differencing of convection term in a two dimensional problem with f v u/Lx . 4.112 1 1 x CX x 1 (l+f) (1+ x ) 1+f 1 1 f 0 x V where f =y u Ax Fig. Stability condition in the plane (Rx.18.
n+l n t n+l n+l n+l n+l n+l S i + u i +xu i i= D i+l D i Lx 2 ii (4. parison between the analytical result in Fig. 4.20. 4.113 Then the number replaced with on the C axis in Fig.17 should be Table 4.102) dx > .100) For a general two dimensional problem.17 and numerical experiments.18. 4. (4. d >  greater Cx (4.99) The stability condition is shown in Fig. 4.101) (4. and it can be seen that the implicit scheme is unconditionally stable if the diffusion constant is than . 4. the stability condition is found to be the following. 4.4.2 Implicit Scheme The following Eq.2 shows the com 1 l+f in Fig.Cx d > .19 and Fig.99) (4.C Y y .uLx. 4.4.99 is a fully implicit finite difference equation for a general one dimensional problem and the Von Neumann analysis will be applied in the same way as in the explicit scheme.
019 Numerical Experiments 0.499 0. R x 0.333 833.147 0.333 4.380 0.833 8.2.403 0.260 0.167 2.3 8.180 0.780 0.060 Table 4.540 0.120 0.660 0.660 0.333 83.629 2.120 0.114 Maximum Stable C. Comparison of the stability conditions in terms of the cell Reynolds number by Von Neumann analysis and numerical experiments in a two dimensional explicit scheme with convection donor cell differencing of term .083 Von Neumann Analysi s 0.720 0.488 0.
C x 2x x 2x C x dx 1(C2+C 2xx ) Fig.dx) for an implicit scheme with donor cell differencing of convection term in a one dimensional problem . 4.19.115 d =  x d =. Stability cndition in the plane (Cx.
116 Cx=1  2 Rx 2 1 0 R x R C x R +2 Fig. Stability condition in the plane (Rx.Cx) for an implicit scheme with donor cell differencing convection term in a one dimensional of problem .20. 4.
The Von Neumann analysis of the first formulation in Eq.4.+d )] C ) {[1+(lcosex +dx)) +[fsinex }{[l+(1cosy) (+dy) + Y[siney } (4.117 4. 4.jto be less than or equal to one. the following condition should be satisfied.63. One of the formulations has both the x and y direction components in each step. C d x>2 (4. treating them implicitly one by one at each fractional step and the other formulation has only the x or y direction component treated implicitly in each step.4.103) jsj <1 In order for I.3 ADI Scheme There are two possible formulations of the Alternate Direction Implicit (ADI) scheme given in Eq. 4.63 gives the following stability condition. 2 _C C 2 {[(cos )(+d X))] C [sine I2} {(1cos6 C 2 )(.104) .
C d x > 2 2x (4. 4. . 4. C d y>  2  4 (4.106) C d y  >  ~ 2 (4.105) The stability condition of the second formulation in Eq.63 can also be given by the Von Neumann analysis as follows.107) Therefore the second formulation of the ADI scheme in Eq.118 The same condition should hold for the ydirection.63 has the same stability condition as the fully implicit scheme.
1 and Fig. occurs as a result of the buoyancy force due to density .119 CHAPTER 5 RESULTS Some calculations are performed to validate the physical models and numerical schemes in the previous chapters. Natural convection occurs due to the heat transfer between the obstacle and containment air. 5. the ADI scheme is explored to use a time step size larger than the Courant limit to save the overall computational time. Finally.1. 5.1 Natural Convection Results The containment is modelled as a two dimensional rectangular compartment with an obstacle as a heat sink. energy convection and heat transfer models are considered and the resulting flow fields are given in Fig. density is determined by temperaSince the natural convection of air ture and pressure. 5. For example. Updating of the reference state. The physical models are tested in a natural convection problem and the numerical schemes are tested in a simple geometry to compare numerical solutions with exact solutions. 5.2.1 Updating of the Reference State The state of air is determined by any two independent properties.
Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 10. 0.5 ft/sec h = 1. = 677.120 p ).1. = 680F air T = 580F Fig.8537x10tu/ft20 F sec (Cp) = 1. 5.0 Btu/lbm 0F (Q ob = 208.8 sec .4 sec AT = 100F T .29 Ibm A= 2 ft2 T.
0 Btu/lbm OF (c )b = 208. 5.4 sec 1 0 AT = 10 F T air = 680F . 5 ft/sec h = 1.8537x10 tu/ft2 F sec p)ob = 1. = 677.121 > 0. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 62. T ab = 580 F Fig.2.5 sec .29 ibm (K 2 A= 2 ft T.
time Pressure of the fluid at old and new steps. not on the pressure because of a high modulus of elasticity. The equation of state of an ideal gas is given in the following. new Pnew RT new (5.total Pold Told new old Qcell cv Mf. Told' Tnew: Temperature of the fluid at old and new time steps.1) T Pnew = old + Ttotal c v M f. temperature and pressure are equally important parameters in determining the flow field. however. Pold' Pnew: .122 gradients. The density change of liquid. depends primarily on the temperature.cell where Pold' Pnew: Density of the fluid at old and new time steps. c : Specific heat of the fluid at constant volume.
Mf. 5. An experimental correlation is given in Eq. 5. convection energy transfer should be in terms of enthalpy instead of internal energy. The value of k is 1. Qcell total : Heat input to the fluid in one cell and in the total domain. .total : Mass of the fluid in one cell and in the total domain.2 in terms of the dimensionless numbers. but they may also be based on the vertical height of the obstacle. The energy convection will be underestimated by a factor of 1.0 Monatomic. The VARR calculations 121] have shown that the magnitude of velocity can be affected by a factor two by updating the reference pressure at every time step. diatomic and polyTherefore. k = c /C for liquid and 1.1.1.4 when internal energy is used for enthalpy in air flow.123 Mf.2 Energy Convection Another difference between liquid and gas is the ratio of the Specific heats. . The Nussett and Grashof numbers here are based on the axial length of one computational mesh.cell.3 Heat Transfer Modelling The heat transfer to the obstacle is modelled as a natural convection from a vertical surface.. 5. the atomic gases have different values of k.4 for air.
109 > 109 C 0. Gr = 9.2179 x 10 B = 1.7134 x 108 Nu = 89.0763 ibm/ft 3 g = 32.124 Nu = C(GrPr) n ( 5.021 n 0.876 The heat transfer coefficient is given as.555 0. L = 2 ft AT = 500 F p = 0.2 ft/sec' 5 V = 1. . h = 1.4 lbm/ftsec Btu/ft 2 F.708 (air) GrPr 105 .923 x 103 Therefore.25 0.8537 x 10 . This may be a typical value for the natural convection heat transfer in the containment without any phase change.sec .3) Pr = 0.4 The following data are used to calculate the heat transfer coefficient at the obstacle wall.2) 3 2 Gr = gS(AT)L p 2 (5.
4) where CQ is the source term in the energy equation of fluid.1 and Fig. T. 5. Obstacle density. The time constant. Specific heat at constant pressure. Heat transfer coefficient between the fluid and obstacle.1. can be obtained from heat balance between the obstacle and fluid as follows.5) where M A : Mass of one computational mesh.4 sec. T = 677.2. If the flow is turbulent. pf Pob: c h : Fluid density.4 Turbulence Modelling Laminar flow is assumed in obtaining the results in Fig. (Mc )ob (MC) obf hApo b ob f (5. 5. the heat . CQ = Cpob (TT) (5. Heat transfer area between the obstacle and fluid over one computational mesh. Therefore. 5.125 The VARR input [21) requires the time constant of the heat transfer instead of the heat transfer coefficient.
3(A) is parallel with the mesh orientation so that no crossflow diffusion occurs. The corrective scheme is tested in recirculating flow problems for mesh point and mesh interface implementations. The results in Fig.3. 5. 5.3(B) is skewed at 450 to the mesh orientation and both truncation error and crossflow diffusions are included in the solution error.to the obstacle will be enhanced and the diffusion transport of momentum and energy will also be enhanced in the containment air. steadystate problem in Fig.6 and Fig. Figure 5. The skew differencing and corrective Huh's formula schemes are tested for various problems. 5. The flow in Fig.126 transfer. The flow in Fig. 5.2 Numerical Diffusion The truncation and crossflow diffusion errors are compared in a simple geometry where an analytical solution can be obtained.5 show that the numerical solutions is close to the analytical solution and the truncation error can be neglected.4 and Fig.1 Truncation Error and Crossflow Diffusion The truncation error diffusion is compared with the crossflow diffusion in a two dimensional. 5.7 show that excessive numerical diffusion has occurred due to the . is compared with De Vahl Davis and Mallinson's in predicting the magnitude of crossflow diffusion. 5. 5. 5.2.
0 ay AB=BC=5.0 u = v = 7.0 a= 1. for evaluation of numerical diffusion .07 (B) =0.0 u = 10.0 B T specified Fig. 0 ST By =0. 5.0 BT S=0.127 (A) .0 x BT  @x a( aT + 2 2 a2 ay ) Th= 10. Problem geometry for the two cases.0. (A) parallel flow and (B) cross flow.0 (A) (B) AB=BC=5.0 B T_ 0.3.
5.128 Analytic solution Numerical solution with Donor Cell T 10. Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 10x10 meshes along the line CD in Fig.3(A) . spatial coordinate Fig.4. 5.
~ I~. Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CD in Fig. 5.~ 129 Analytic solution Numerical solution with Donor Cell C spatial coordinate Fig.^XI ill I~LX_L~LI.5. 5.3(A) .__~ .__LL l.~l r.
5. Comparison of the analytic solution.3(B) . 5. analytic solution with increased diffusion constant and numerical solution with donor cell differencing convection term in 10x10 meshes along the of line CA in Fig.6.130 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Df cf T 10 spatial A coordinate Fig.
analytic solution with increased diffusion constant and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CA in Fig.7. 1/ 5.. 5.3(B) .131 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Dcf T 10 / I / / ./ Fig.3(B) s' / line CA in . 5. . Comparison cT the analytic solution.J 0 I' C spatial A coordinate Fig.
2. 5.7 show that the numerical solution can be reproduced by the analytical solution with increased diffusion constants by the amount given by Huh's formula.9 is the result by Huh's formula and Fig. 5.53 and Eq. 5. 5.7 are for 6 x 6 It is seen that a finer mesh gives a better Since the crossflow diffusion constant is pro portional to the mesh spacing.12 for a simple pure . Figure 5. 4. 5. more diffusion has occurred in Fig.11 and Fig. Eq. Since the corrective scheme is based on the validity of those correction formulas.3 Skew Differencing Scheme Two skew differencing schemes. 5.6 are for 5. 10 x 10 meshes and Fig. De Vahl Davis and Mallinson's and Huh's formulas give the same crossflow diffusion constants for the case. It can be seen that the former result is much better than the latter. Figure 5.5 and Fig. 5. 5.54.10 by De Vahl Davis and Mallinson's for the problem in Fig. The problem geometry in Fig.2. are tested in Fig. solution. it is necessary to test them for various cases. 5.8.4 and Fig. Raithby's and Huh's. 5. Figure 5. 4.7 than in Fig.810. 6 = 6 = 450. meshes. 5.6.132 crossflow diffusion. 5.2 Validation of Huh's Correction Formula The crossflow diffusion can be predicted by Huh's formula.6 and Fig.8 has the arbitrary angles of 6 and 61 such that 6 = 600 and 61 76.
335 Huh A e1 = 76.810 8 = 600 8 Meshes 24 Meshes Fig.7674 DH u h = 3.0 Ax = 0.6603 DDM= 0.8.133 Yo 2 T=0 D a = 1.0267 u = 5. Problem geomctry with arbitrary angles of 6 and 81 to compare De Vahl Davis and Mallinson's and Huh's formulas for prediction of crossflow diffusion constant . 5.2406 by = 1.0 v = 8.
5. 5 0 spatial coordinate Fig. Comparision of the numerical solution with donor convection term and cell differencing of analytic solution with increased diffusion constant. (D+Dcf). by Huh's formula along the line BD in Fig.9.134 Analytic solution with D+Dcf Numerical solution with Donor Cell T 10. 5.8 .
solution (D+DDM). 5. Comparision of the numerical solution donor cell convection term and analytic differencing of with increased diffusion constant.10. 5. by De Vahl Davis and Mallinson's formula along the line BD in Fig.135 Analytic solution with D+DDM Numerical solution with Donor Cell T 101 spatial coordinate Fig.8 .
75 25 U 12.136 Donor cell 100 100 100 100 100 93.25 18.5 68.25 65.38 50 31. Comparison of the true solution and solutions by donor cell scheme.25 0 Skew differencing by Huh (True solution) 100 100 100 100 50 0 0 0. 5.63 87. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with e = 45* .75 81.11.75 75 50 U Skew differencing by Raithby 50 100 100 100 100 50 100 100 100 50 0 100 100 50 0 0 100 50 0 0 0 50 0 0 0 0 50 34. 0 100 100 100 100 100 50 100 50 0 50 0 0 V Fig.5 0 6.
11 3.3' 40.25 1.52 1.9 33.5 0 0 63. 5.56 1.01 55. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure covection problem with e = 63.6.59.91 31.25 37.83 72.25 53.70 4.50 2.47 23.8922.2! 90.86 16.67 5.91 71.3: 11.E 21 0 12 True solution 3.25 0 0 0 .12.99 10.22 9.5C 100 100 50 Fig.11 0 11.71 20.430 0 50 0 Comparison of the true solution and solutions by donor cell scheme.43 7.137 Donor cell 100 100 100 100 50 .74 38.430 .85 0.96 17.3E 21.88 81.6 0 Skew differencing by Huh 100 90.A Skew differencing by Raithby 100 100 100 100 96.5E 25.1 6. 62.26 80.13 0 0 100 100 100 100 100 100 50 25 0 50 25 0 0 0 0 0 0 0 0 0 0 0 0 6.33 70.
but less than inthe donor cell scheme.2. 5. both of them are conservative in a unidirectional flow.13 and Fig.14 show that the analytical solution can be reproduced by Huh's skew differencing scheme for the case.4 Corrective Scheme The prediction formulas of the crossflow diffusion constants are validated by showing that the numerical solution can be reproduced by the analytical solution with the appropriately increased diffusion constant.3. 5. method while Huh's scheme is Lagrangian. Some crossflow diffusion has occurred in Huh's scheme. Raithby's scheme is an Eulerian Therefore. 5. The details of those schemes are Figure 5. the However. Fig.138 convection problem. at a 450 angle with respect to the mesh orientation and shows that the donor cell scheme introduces appreciable crossflow diffusion while Raithby's and Huh's skew differencing schemes give the true solution.3. to obtain the true solution by subtracting the additional crossflow diffusion . however.430 angle and neither Raithby's nor Huh's scheme can reproduce the true solution. The purpose of the corrective scheme is.2. 8 = 1e= 45*. Figure 5. former is conservative while the latter is not.12 has a flow at a 63. It is indicated that Raithby's scheme may give unphysical results in problems with steep gradients of the quantity under consideration.1 and 4.11 has a flow given in sections 4.
5. 5.13. Comparison c f the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 10x10 meshes along the line CA in Fig.139 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh C A spatial coordinate Fig.3(B) .
14.140 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh 10 5 0 spatial A coordinate Fig. 5. Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 6x6 meshes along the line CA in Fig.3(B) . 5.
The corrective scheme gives a physically reasonable solution. Figure 5.17 and Fig. Figure 5. although not identical to the true solution. Figure. 5. 5. are introduced in Chapter 4. In Fig.19 is a hypothetical 3 x 3 recirculating flow field where the inlet boundary values are specified on the left and bottom surfaces.141 constant from the total diffusion constant of the finite difference equation.15 numerical solutions are given for a simple pure convection problem with various diffusion corrections. The full correction of the crossflow diffusion gives the true solution without any numerical diffusion error. 5. 5. Fig. The corrective scheme is also tested in the problem geometry of Fig.430. 5. mesh point and mesh interface.3(B). donor cell solution and the solutions by mesh point aand mesh interface implementations. 5.20 shows the diffusion corrections of the mesh point and mesh interface implementations at each interface. Two implementation strategies of the corrective scheme. Fig. Both of them are tested in a recirculating flow problem with pure convection.21 gives the true solution.16 has a flow with 8 = 63.18 show that the crossflow diffusion can be eliminated by the corrective scheme and that a finer mesh spacing always gives a better solution. The mesh point implementation gives an unphysical solution . The flow is at a 45* angle with respect to the mesh orientation.
75 6.71 0.08 5.01 7.25 0 5.59 7. Donor cell solutions with various diffusion corrections for a pure convection problem with e = 450 .41 u=v=l Ax=Ay=l D =D =0.142 D=0.0 8.1 10 8.0 2.29 5.88 10 5 0 2.0 10 7.99 D=0.5 (True Solution) 9.5 x y Fig.12 5.0 2.0 2.71 5.15.13 7.5 0 1. 0 V ________ D=0.92 5.5  3. 5.3 D=0.0 1.29 0.88 5 4.
73 10 7.37 3.59 1.10 9.96 3. 0 D2 3 0.11 0.33 1.07 10 3.56  2.16. True solution and donor cell solutions with and without diffusion correction for a pure convection problem with e = 63.430 u=l.93 10 5.91 0.11 6.11 True Solution V 63.430 . v=2 x= y=l 2 D =D = x y 3 Fig.143 D=0.82 0.07 0.41 0. 5.04 4.07 2.
5. numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig.17. Comparison of the analytic solution. 5.144 Analytic solution Numerical solution with donor cell Solution by corrective scheme 10 A spatial coordinate Fig.3(B) .
5.3(B) . 5.145 Analytic solution Numerical solution with donor Solution by corrective scheme 10 5 0 spatial coordinate Fig. numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig. Comparisoin of the analytic solution.18.
19. Simple recirculating flow field for test of the implementation strategies of the corrective scheme .146 0 Ofr 3 ~LC~' / 10 1 1 5 0 1 3 3. 5. 3 ) 2 3 2 Lx = Ay = 1 Fig.
20.25 mesh point implementation 0.5 0.125 0.2 0.75 0.147 0.75 0 0 0.2 0 0 0 mesh interface implementation Fig.625 0.19 . 5.75 1. 5.5 0 0 5 0 0 .625 0 .75 1. Diffusion coastant corrections at each interface for the mesh point and mesh interface implementations of the corrective for the recirculating flow field given in Fig.25 1.75 1.
148 True solution Donor cell mesh point implementation 5.94 9.51 10 5 0 Fig. donor cell solution and solutions by the mesh point and mesh interface implementations of the corrective scheme for the recirculating flow field given in Fig.83 6.10 8.21 mesh interface implementation 0.18 7.21 7. Comparison of the true solution.08 9.86 12.82 8.21.10 9.26 8.19 . 5.75 6. 5.16 6.16 11.
Another test calculation is done for a two dimensional.26. 5. Figure 5. respectively.2 kg/sec.149 while the mesh interface implementation gives a physically reasonable solution.25 and Fig.22 shows the flow field in the containment and more detailed information about the flow field is given in Fig. The steam concentration distribution at the beginning is given in Fig.28 is not a reasonable solution while the mesh interface implementation result in Fig. The result in Fig. Figure 5. 5. flow field is used here to test the mesh point and mesh interface implementations of the corrective scheme. There was initially air in the containment and steam is introduced in the source mesh at the rate of 0.27 is physically reasonable and even better .23. Since the steam concentration distribution and flow field are given at the start of the calculation. 5.3 sec with the corrections in Fig. 5. 5. 5. 5. 0.24 in which the source mesh has the highest steam concentration. After a while a flow field is set up in the This containment by the input mass and momentum of steam.25 and Fig.26 show the diffusion constant corrections of the mesh point and mesh interface implementations and Fig. rectangular containment with an arbitrary recirculating flow field.19.28 show the steam concentration distributions after one time step. the steam concentration for the new time step can be calculated in an explicit way. 5. 5.27 and Fig. 5. The true solution cannot be reproduced because some of the information about the flow field has already been lost in Fig.
0m Fig. 5. Recirculating flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the convective scheme .5m Ay = 1.150 Ax = 0.22.
4870 0.5893 0.5174 1.2474 04 1379 0.2435 0.1422 2148 0. Detailed flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .23.1139 0 .151 0.2216 0.1385 0. 5.456 0.2460 0.207 2.3957 0.4920 0 .0348 0.0937 0 1. 3563 0.5 m Ay = 1 m Fig.3 095 1.1276 0 . 2214 velocity [m/sec] Ax = 0. 045 0.34 0. 0648 1.3074 0.2629 0.2190 0.1229 0.3253 0.531 0. 0735 0.7126 0.0492 0 .2159 0. 861 1.93 2. 5955 SOURCE 0.0127 0. 1317 0.
1750 0.24.0136 0.1244 0.0819 0.0179 0.1085 0.5m by = Im Source: steam concentration in Ikg/m 3 0.0676 0.0620 0. Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .1712 0.0007 0.1592 0.0298 0.0358 0.0011 0.2 kg/sec of steam Fig.1178 0.0095 0.1859 SOURCE 0.1710 0.152 0.0346 0.0305 0. 5.1420 0.1266 0.0031 Ax = 0.0049 0.0002 0.
1854 0.1 389 0. 5.25. 2422 0.C209 0.1694 3106 0.0 278 0.1279 0.0311 0.0224 0.3262 0.0311 0.stant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .0255 SOURCE 0.2740 0. Diffusion co.0600 0.0590 0.0300 0.1014 258 0. 0.0 183 Ax = 0.. 0.1763 0. 0.0193 0. 0488 0.0731 116 0.0794 0.0937 0.0348 0.1451 0.1178 0.1452 0.0440 0.1583 0.5m by = Im Diffusion constant in Im2/sec] Fig. 0152 0. 0227 0.0 362 0.1701 0. 0086 0.153 0.0253 0. 0396 0.
0 923 0 0. 0579 0.0192 0 0. Diffusion constant corrections of the mesh interface implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .154 0 0 0. 0240 0.0215 0.0860 0.0518 0 0 0.0632 0.0 257 0.2134 0.4732 0.1026 0 SOURCE 0 0 0.0479 0.0534 0.0443 0.0 76 0.0287 0.1149 0 0 0. 5.9769 0 0 0. 0706 0.2529 0.3694 0.5m by = lm Diffusion constant in Im2/sec] Fig.2317 0 0.26.0 111 0 Lx = 0.
1733 0.0833 0.0195 0.1706 0.0735 0.1321 0.0947 0.3 sec.0342 0.0016 0.0233 0.1623 0.1243 0.155 0.0448 0.1847 SOURCE 0.0107 0.0068 0.0342 0. 0.27.1711 0. 5. with donor cell differencing convection term without any diffusion of correction .2 kg/sec of steam Fig.0003 0. Steam concentration distribution after one time step.0011 0.1135 0.0465 0.1363 0.1475 0.0046 Lx = 0.5m Ly = Im steam concentration in [kg/m 3 ] Source: o.
0815 0. Steam concentration distribution after one time step.1747 0.1911 0.0937 0. 5.3 sec.0250 0.0287 0.0005 0.28.156 0.0465 0.0010 0.0229 0.0042 Lx = 0.0068 0.2 kg/sec of steam Source: 3 Fig.5m Ly = Im steam concentration in Ikg/m 0.1797 0.1334 0.2086 SOURCE 0.1646 0.1240 0.1247 0.0029 0.0433 0.1493 0. with mesh point implementation of the corrective scheme .0355 0.1729 0.0709 0. 0.0092 0.
1342 0.1847 SOURCE 0.0007 0.1487 0.0039 ax = 0.0295 0.0481 0.29. 0.0722 0.2 kg/sec of steam Fig.0228 0.0424 0.0058 0.0003 0.1310 0.157 0.5m by = Im steam concentration in [kg/m 3 Source: 0. 5. Steam concercration distribution after one time step.1794 0.1243 0.0100 0.3 sec.0360 0.0947 0.1738 0.1628 0.0192 0.0010 0.1756 0.0815 0.1112 0. with mesh interface implementation of the corrective scheme .
5. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable.0 and the inlet boundary value of 4 is. it has the maximum time step size that can give a physically reasonable solution. Figure 5.158 than the donor cell solution.3 ADI Solution The ADI scheme is tested to use a time step size larger than the Courant limit. In the meshes at the top of the source mesh.31 shows the profiles of 4 along the dotted line for the time step size of 0. The ADI and explicit schemes give almost identical results during all the transient.30 there is an inflow from the bottom surface at the velocity of v=1. The results of the ADI and explicit schemes are compared in the problem geometry of Fig. when the Courant limit is 1 sec.0. =10. There is a source of 4 on the right boundary over two computational meshes and the profile of Q along the dotted line is affected by that source. which is exactly the . 5.32 shows the profiles of 4 at the same loca tion for the time step size of 1 sec. 5. In Fig. the steam concentration should decrease as the flow goes up to the ceiling because most of the steam introduced in the source mesh goes upward and gets distributed by convection and diffusion.5 sec. Figure 5. It is found that a wise use of the ADI scheme can reduce the overall computation time in comparison with the explicit and implicit schemes.30.
Problem geometry for comparison of the explicit and ADI solution schemes .~ ~ILI_~_~LI*L~l*IIIU*XI 1 9 IB Source of ' 'A Profiles of 4 = 10.0 c= 1.30.YXI~^.0 on this line are compared. 5.1I I _~ 1_.0 Ax = Ly = 1. v = 1.0 Fig.
0A 12.0 11.0 coordinate 11.5 .0 10. 5.0 10.31. 0A Explicit ADI 10.01 12.0 11.0 10.0 Fig.0 11. 5.0 A iatlai B 10.0 12.0 12.0 11.0 0.0 10. Comparison of the explicit and ADI solutions for along the line AB in Fig.0 11.0 sec for Lt=0.5 sec 12.160 12.30 the profile of sec and the Courant limit of 1.
0 10.0 sec and the Courant limit of 1.30 for At=1.0 10.0 Fig.0 11.0 10.161 1 sec 12.0 4 sec 11. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig.0 11. 5.0 12. 11.0 . 5.0 12.0 16 sec 20 sec 12.0 Explicit ADI 10.32.0 11.0 sec . 12.0 10.0 ~ spatial coordinate 12.
. 5. there is some deviation between the two results. No crossflow diffusion corrections are involved in the results of Fig.33. 5.162 the Courant limit.32 and Fig. 5.31. Fig. but it damps out rapidly and the ADI scheme gives almost identical results with the explicit scheme.33 shows that the ADI scheme can use a time step size up to five times the Courant limit. Fig. The deviation for the time step size of 5 sec is expected to damp out as the calculation goes over to the next time steps because of the unconditional stability. Initially. 5.
163 12. 5.0 At =0.33. Comparison of the ADI solutions for the profile of at t=10.30 with different time step sizes . 5.0 sec along the line AB in Fig.0 t spatial coordinate B Fig.5 sec ec 11.
and turbulence equations are solved separately using that flow field. The .2 Numerical Schemes (1) There are two numerical diffusion sources. Crossflow diffusion occurs due to the donor cell treatment of the convection term in a multidimensional problem. The con tinuity/momentum equations are decoupled from the scalar transport equations and solved by the SMAC scheme to obtain the flow field. or superheated steam as may be appropriate. (2) The models of laminar and turbulent diffusions in Chapter 3 may be adequate for predicting the hydrogen transport in the containment. 6. The thermal equilibrium is assumed and the phase change occurs to maintain 100% relative humidity.1 Physical Models (1) The solution scheme presented in Chapter 2 has been adequate for modelling the slow mixing stage in the containment after a lossofcoolant accident. in the finite difference donor cell treatment of convection. truncation error and crossflow diffusion. energy.164 CHAPTER 6 CONCLUSION 6. The mass diffusion. The total diffusion con stant is the sum of the laminar and turbulent diffusion constants.
but not always. (2) Two types of approaches. The truncation error diffusion occurs in the flow direction while the crossflow diffusion occurs in the diagonal direction of neighboring mesh points. . most of the numerical diffusion error in a multidimensional. Therefore. ~unx.165 effective diffusion constants of the truncation error and crossflow diffusion are of the same order of magnitude. It is conservative and an explicit scheme can be used without affecting the simple solution structure. recirculating flow problem is due to the crossflow diffusion. Therefore. The gradient of the scalar quantity under consideration is usually small in the flow direction in comparison with that in the direction normal to the flow. the corrective scheme is generally preferred to the skew differencing scheme. The skew differencing scheme gives good results for some problems. convection dominant. The corrective scheme is based on the fact that the additional crossflow diffusion can be predicted theoretically for every mesh at every time step. skew differencing and corrective schemes. It gives unphysical results for most recirculating flow and coarse mesh problems. have been tried in order to eliminate the numerical diffusion. The inclusion of corner points complicates the matrix structure and a fully implicit scheme should be used for maintenance of stability.
4. (4) The Von Neumann stability analysis is applied to various finite difference forms of a general conservation equation.166 (3) Two implementation strategies. it has its own limitations due to the following factors. are identical in a unidirectional flow. the physical constraint in section 4. are tried for the corrective scheme. mesh point and They mesh interface.2 imposes a maximum time step size that can give a physically . i. First. (5) The maximum time step size is limited by the Courant The Alternate Direction condition in an explicit scheme. The graphical method is used to obtain the stability condition in terms of the Courant. The results of various sample problems show that the mesh interface implementation of the corrective scheme always gives a physically reasonable solution with negligible numerical diffusion error. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable. The results of numerical experiments are found to be consistent with the Von Neumann analysis.. dx and Rx . Implicit (ADI) scheme can be used to increase the time step size and decrease the overall computational effort. The mesh interface implementation is always preferred to the mesh point implementation in a recirculating flow problem.e. diffusion and cell Reynolds numbers. using the characteristics of a quadratic equation. Cx.
. For example. Second.167 reasonable solution. a steadyThe state solution may never be reached due to the fact that one time step is composed of a few fractional steps. the ADI scheme is a fractional step method. asymmetry of the given problem may give different results according to the sweeping sequence.
geometry (e.. to eliminate the crossflow diffusion. previous time step information. A criterion for the flow regime should be developed in such a form that it may be implemented into a computer code. The modified kE model in a low Reynolds number flow may be a guide in this approach. transitional and laminar flow regimes. The important parameters may be the velocity. time step size. skewdifferencing and corrective schemes. velocity gradient. They may still be improved further to get an accurate solution in a transient. The flow regime criterion may be replaced with a model that covers all the turbulent. The model should show proper limiting behaviors as the Reynolds number goes to infinity (turbulent) and zero (laminar).1 Physical Models Since both the laminar and turbulent flow regimes are possible in a complicated geometry and flow field. multidimensional recirculating flow problem. it is necessary to predict the flow regime in order to use appropriate physical models. distance from the wall).168 CHAPTER 7 RECOMMENDATIONS FOR FUTURE WORK 7. . 7. etc.2 Numerical Schemes There have been two approaches.g.
Skew differencing scheme 1. More tests and validation calculations are required in recirculating flow problems. 2. It may be possible to develop a reasonably simple form that may or may not include the corner points. . The solution in Appendix A may be a useful guide in this work. The conservative form of the skew differencing scheme is complicated and time consuming in comparison with the corrective scheme.169 Corrective scheme The corrective scheme has been successful by the mesh interface implementation strategy in the scope of this work. ADI and implicit schemes although only the implicit scheme has been used up to now. The stability of the skewdifferencing scheme is open to question in the explicit.
24.E. 34. "Solution of the Multidimensional Compressible NavierStokes Equations by a Generalized Implicit Method. Cebeci. "A New Finite Element Formulation for ConvectionDiffusion Problems. S. Amsden. and McDonald. F. F. T. "A Simplified MAC Technique for Incompressible Fluid Flow Calculations.R.. A.B.H." Journal of Comp. R." HEDLSA2682 (October 1982)... Phys.. Transport Phenomena. Stewart.R.N. 233241 (1975).R.H.D. et al. C. "A Finite Difference Scheme for the Incompressible AdvectionDiffusion Equation. Baliga. thesis (1979).. Academic Press (1981)..R. "Transient Effects in Turbulence Modelling." Journal of (1968).G. H. and Harlow. and Whitelaw.A. E. 6." MIT Ph. Boyle." NUREG/CR1148 (EGG208). 94110 Amsden.V. 5473 (1980).. A.A.. and Harlow. Broadus." Journal of Comp. Bird. J.A. John Wiley & Sons (1960). "On the Structure and Use of Linearized Block Implicit Schemes. Engineering Calculation Methods for Turbulent Flow. "The SMAC method: A numerical technique for calculating incompressible flows." Comp. Phys. B. Phys. (April 1980)." Numerical Heat Transfer 3. Comp. Bloom. Briggs. and Patankar.User's Manual. and Harlow. Briley.. 332325 (1970). W.H. P. 393409 (1980). W. W. H.. D. D." Journal of Comp..R. F. G. 6. 372397 (1977). "BEACON/MOD 3: A Computer Program for Thermalhydraulic Analysis of Nuclear Reactor Containments . "Transport of Turbulence in Numerical Fluid Dynamics.H. Bradshaw." LA4370 (May 1970).. A. 3. et al. Briley. Phys.. and McDonald..R.170 REFERENCES [1] Amsden. and Lightfoot. Methods in Applied Mech. [2] [3) [4] [5) [6] [7] [8) [9) [10) [11) [12] . and Eng. "Hydrogen Distribution in a Containment With a HighVelocity HydrogenSteam Source.
Method for Convection in Numerical Fluid Dynamics. Methods in Applied Mech." LA6236 (July 1976)." MIT M.B. Nielsen. Varadi. Phys.G. G..D. P." Comp.S. De Vahl Davis. and Lewis.. "Buoyancy Effects in Mathematical and Numerical Turbulence Modelling.D. N. Clark. [14] [15) [16] [17] 118] [19] [20) [21] [22] 123] [24] [25) . 95107 (1982).. Jr. thesis (1981). Cook. 27. 1344 (1981). Hirt. Keller.. and Design 69. "Studies of Numerical Methods for the Plane NavierStokes Equations.. Crowley.S..H. 127137 (1982).. L.. M." CRBRPARD0106." Journal of Comp. 32.D. J. Williams." Nuclear Eng.H. S. G. J. Chen..I.. J. Cheng. 471484 (1967). M. 1. Vol.A Computer Program for Calculating Timedependent Turbulent Fluid Flows with Slight Density Variation. and Eng.I. "Study of Nonhomogeneous Conditions in Reactor Containments Due to Separation Effects. "Accuracy of Difference Formulation of NavierStokes Equations. S. C. Vol." Journal of Comp. of Fluids Supplement 2 (1969).F.K. and Design 69. L.. "SOLAICE: A Numerical Solution Algorithm for Transient Compressible Fluid Flows. Phys. thesis (1977).. 1 (Nov... 2943 (1976). Cebeci. Cloutman.J. 1976). S.D. Chang.C.W. "VARR2 ." The Phys. "An Evaluation of Upwind and Central Difference Approximations by a Study of Recircul ing Flow.. Academic Press (1964).P. and Ramshaw. and Nakayama.D. J. "SecondOrder Numerical Advection. B.L." Computers and Fluids." MIT Ph. Covelli. P.B. thesis (1980). K. and Mallison. and Hansen. 4. Hirsh. and Romero. "Simulation of Containment Cooling with Outside Spray After a Core Meltdown." Nuclear Eng.. R." MIT Ph. "Coolant Mixing in the LMFBR Outlet Plenum.. Y. 7179 (1979). "Tensor Viscosity Dukowicz. W.. G. Chang. "Numerical Methods of Reactor Analysis. H.171 [13) Carbajo.. T.. "Comparative Analysis of Numerical Methods for the Solution of NavierSTokes Equations.
C. and Marohar.. of Fluids Supplement 2 (1969)..A. Vol. 2.." Int.M. Phys.." The Phys." The Phys.E. J. Turbulence.O.. R. "A Numerical Fluid Dynamics Calculation Method for All Flow Speeds." Journal of Comp. and Golay. M.P. "Heuristic Stability Theory for Finite Difference Equations. Hamza." Journal of Comp. C... Harlow.. 3..W. "Improved Form of the Low Reynolds Number kc Turbulence Model." MITEL 81020 (June 1981).A. Harlow. "Buoyancy Induced Fluid Motions Characteristic of Applications in Technology . B.W. Curtiss. "Numerical Calculation of Almost Incompressible Flow. of Fluids Supplement 2 (1969).W. [38) . New York (1954).. B. Vol. Phys. and Bird. Hirschfelder. Hinze. Methods in Eng. 528 (March 1979)." The Phys. "A Comparison of Hybrid and Quadratic Upstream Differencing in High Reynolds Number Elliptic Flows. J..H.. 18. [27] [28] [29) [30) [31] [32) [33) [34] [35] [36) [37) Hirt.. Gupta. of Fluids. 2nd ed.C. R.H. Gebhart. Hirt.A.. McGrawHill Book Company.B. 8. A. Phys." Journal of Comp. Humphrey. 101. A. R. "Behavior of Buoyant Moist Plumes in Turbulent Atmospheres. 8093 (1968).O. and Eng. M. (1975). and Launder. F. and Amsden. 557573 (1980).. Methods in Applied Mech. 1952 (1975). T." Journal of Comp.A. "Practical Investigation of Convective Difference Approximations of Reduced Dispersion.. and Amsden.E. Harlow. 15. F. 309312 (March 1975). A.. Molecular Theory of Gases and Liquids. J. "Computer Studies of Timedependent Turbulent Flows. "On the Use of Central Difference Scheme for NavierStokes Equations.H. "Numerical Calculation of Multiphase Fluid Flow. C. 17. and Amsden. Hoffman. F. 197213 (1971). Phys.. 29." Journal of Fluids Eng.172 [26] Fromm. Wiley. G. Vol.. J. J." Comp.. Han.The 1978 Freeman Scholar Lecture.F. 8195 (1981). 339355 (1968)... for Num.H.
H.W. Langer. Lam." Journal of Comp. 1981). 537566 (1975). Keenan. 293312 (1980). M. B.K..E. "The Prediction of Laminarization with a Twoequation Model of Turbulence. B." Comp. Launder.. thesis (1983).. D. and Spalding.H.G. Reece. 16. Steam Tables. Jones. 148) [49) .P. W. G. "Progress in the Development of a Reynolds Stress Turbulence Closure.D. K. Methods in Applied Mech. V..V.E. Jr. Vol.. Vol. Frankfurt. Heat Mass Transfer. and Golay.. V. Fluid Mech. F.. W.. Vol. "Analytical Modelling of Hydrogen Transport in Reactor Containments. "The Numerical Comparison of Turbulent Flows.. and Eng. 3.173 [39) Houston." MIT Sc.. Cl. translated (October 1980). M. and Launder. 301314 (1972). Heat Mass Transfer." J. Battelle Institute e. et al.A. 23.P. 15." Comp. G. Jones." BF63. J.E.. and Launder. part 3. 68. W. and De Braemaecker. and Eng. "Practical Evaluation of Three Finite Difference Schemes for the Computation of SteadyState Recirculating Flows... 456460 (Sept. 16. Manno. and Keyes..3633. and Rodi.P.E. "Experimental Investigation of Hydrogen Distribution in a Containment of a Light Water Reactor Following a Coolant Loss Accident. Methods in Applied Mech. [40) [41) [42) [43] [44) [45] [46] [47] Leschziner.J. Vol. "Analytical Modelling of Hydrogen Transport In Reactor Containments.." Report submitted to 2nd International Topical Meeting on Nuclear Reactor Thermalhydraulics (March 1982). 221239 (1974). "ADI Solution of Free Convection in a Variable Viscosity Fluid. C. Phys. 11191130 (1973)." Int. B." Int.G. "The Calculation of Low Reynolds Number Phenomena with a Twoequation Model of Turbulence.B. John Wiley & Sons (1978)." Trans. 269289 (1974). 103. Manno. "A Modified Form of the kc Model for Predicting Wall Turbulence. J.. Launder.. M.P. J. J. and Bremhorst. of the ASME. B.
. Methods in Eng.D. 17871806 (1972).. J. "Convergence and Accuracy of Three Finite Difference Schemes for a Two Dimensional Conduction and Convection Problem. K. "A Calculational Procedure for Heat." Int.V. Patankar. Roache. Heat Mass Transfer. Vol. 15.J. of the ASME. Methods in Applied Mech. A.K." The Phys. Reid. Computational Methods in Engineering and Science. 2.. 153164 (1976). Patankar... New York (1958)." Computers and Fluids...D. and Torrance. Roache. G. N. 169184 (1972). 4. 9.V. Phys." Comp. (1977).. and Weiss. P. Rodi. W. R. W. Albuquerque." Journal of Comp.C. 1975). Schonauer. New Mexico (1976)." Int. 541550 (1972). G." Mathematics of Computation. 191206 (1974)." Trans.K. 20 (1966). Mass and Momentum Transfer in Three Dimensional Parabolic Flows.E. [52) [53) [54) [55) [56) [57) [58) [59] [60) [61) [62] [63) . S. McGrawHill. "Upstreamweighted Differencing Schemes and Their Application to Elliptic Problems Involving Fluid Flow. of Fluids Supplement 2 (1969). 75103 (1976). Raithby.B. Hermosa Publishers..O. and Spalding.. "Numerical Experiments with a Difference Model for the NavierStokes Equations (Turbulence Model). S." Comp. 386389 (Sept.. for Num. Pergamon Press (1982).D. P.J. "Convective Difference Schemes... "SkewUpstream Differencing Schemes for Problems Involving Fluid Flow. Methods in Applied Mech. "A Note on the Empirical Constant in the KolmogorovPrandtl EddyViscosity Expression. 9.. "A Critical Evaluation of Upstream Differencing Applied to Problems Involving Fluid Flow. John Wiley' & Sons. W. "On Artificial Viscosity.B. and Eng. G. Turbulent Buoyant Jets and Plumes. Raithby. S. Computational Fluid Dynamics.. Roberts. The Properties of Gases and Liquids... J. Numerical Heat Transfer and Fluid Flow. and'Eng.. Inc. Raithby. Rodi. and Sherwood. 10.174 [50] [51) Nakamura. Hemisphere Publishing Corporation (1980). T. K. Runchal. D.
. and Strong. "A Novel Finite Difference Formulation for Differential Expressions Involving Both First and Second Derivatives. (Jan. 4 (Oct. 153168 (1982). MIT Press (1970). Raithby.H." FATE80114 [73] (Jan. Methods in Applied Mech.Dec.A. 4. "Diffusional Properties of Multicomponent Gases.." SAI74509LJ. 220237 (1976)." Int. "Comparison of Finite Difference Journal of Computations of Natural Convection...G." Numerical Heat Transfer.. . and Hutton. Vol. MatheResearch matical Sciences.J." of the National Bureau of Science ." Comp. C. J. and Eyler. K. 20.. 46. Spalding. G. Y. J.. and Valentin. Smith. 72B. For Num.S. Trent.175 [64] Shin. A. A. 411428 (1980). R.D. H. 3.... "Proposal for a New Discrete Method Based on an Assessment of Discretization Errors. 411428 (1980). 1974). Phys.M. "Simulation of Convection and Diffusion Processes by Standard Finite Difference Schemes and by Influence Schemes. K.. 3." Chemical Engineering Progress. No. Stuhmiller.A. "TEMPEST: A Threedimensional Timedependent Computer Program [72] for Hydrothermal Analysis. No. and Eng.D. Stubley. D.B. G. D. G. M. R.B.... and Lumley. 35. "Numerical Analysis of Fluid Hammer Waves by the Method of Characteristics.L. and Woolner. bulence. Wilke. 1981). Budden. Methods in Eng.B. J. 1968).E. 2.D. A.. "The Numerical Treatment of Advection: A Performance Comparison of Current Methods.D. G. Strong.B. 95102 (1950). Vol.L. Raithby. Tennekes.. "Development and Validation of a TwoVariable Turbulence Model. Stubley. L.W. A First Course in Tur [65] [66) [67] [68) [69) [70] [71] Torrance.R.." Numerical Heat Transfer. 551559 (1972)." Journal of Comp.
176 APPENDIX A ANALYTICAL SOLUTION FOR THE PROBLEM IN FIGURE 5. A.3(B) A steady state two dimensional energy conservation equation without any source is given in the follbwing Eq.1 is2 reduced to the following form.a2T aT 2 + a2 2 2c ax ax 2 y ay (A.1.1 convection occurs in the x direction and diffusion occurs in both x and y directions. Eq. pc u aT = k(a2T + 2aT 2 x2 p x ax ay (A. as follows.2 is simplified further by substitution of 4 for T as follows. 5.1) In Eq..2) Equation A. a= k PCp C =2 u 2A.3(B) and the constants can be grouped into one constant. A. The boundary conditions are specified in Fig. c. Therefore.3) ax . 2 x 2 a2 ay = c 2 (A.
which are functions of x and y only respec (x. Equation A. I d2X + 1 d 2Y 1 = c2 C (A.6) dx dy The two terms on the left hand side of Eq.3 is reduced to the following form.3 can be solved by the method of separation of variables.6 should be equal to some constants because they are functions of x and y only respectively and the sum of them is equal to a constant. A. A.a =  (A.1.5) Then Eq. c 2 X cx 2 = (A.y) = X(x)Y(y) (A.177 where 4 = eCXT (A. ables. A.4) The boundary conditions should also be expressed in terms of 4 as given in Fig.7) d Y c .8) dy where a > c > 0 . The variable 1 is separated into two vari X and Y. tively.
4 = ec T .0 0 = 0 x Fig. Boundary conditions in terms of 4 for the problem in Fig.3(B) where 0 is given by. 5. A.1.= YO T 0 ii ii i h YO + co = 0 4 = 0.178 .
A. Eq. A.10 should satisfy the following relations. A. dY dy y=O Y=O 0 (A. A. the constants in Eq. A.12) (d dx _cX) = 0 x=xO (A.11. The boundary con dition at x=0 will be considered later.7 and Eq.9 and Eq.14) (A.179 The solutions to Eq. 13) In order to satisfy the boundary conditions given in Eq.10) The boundary conditions in Fig.9) Y = C3sin8y + C4cosSy (A. A. C3 = 0 (A. 13.1 may be given in terms of the variables X and Y as follows.16) . X = Cleax + C2 eax (A.15) ny 0 = n C2 _ n+c 2 e nX 1 n (a n 9 C) (A.11) dY dy=Y 0 = 0 (A.12 and Eq.8 can be easily obtained as follows. A.
3 .0 2 A n Th h nr 1 n 1+( a c )e n 1 a +c 2a x . the constants A. nw sin2 nO0 (n=1. ax = A0e + C ax Z A [e n=l n a +c n n 2a x 0 n ea nx + ( a c )e e ]cOSSnY n (A.1 are determined as follows. . . 17) T(x.) (A.2. Th 0 2. the final solution can be expressed as follows.18) n n Since the boundary condition at x=0 should also be satisfied.y) = e cx (x.180 where 2 =c 2 + n (O) n 2 yO =c 2 + 8 2 . Therefore.y) where nr (A.19) .
the time step size in the ADI should be at least two or three times the Courant limit in order to have a gain in the overall computational efforts required for a problem. Therefore.4. When the tilde phase of the SMAC scheme and other scalar transport equations are solved by the ADI scheme.1 and B. The ADI scheme is applied to a two dimeraional momentum equation in the following Eq. the computational effort for one time step is greater than that of the explicit scheme. 181 APPENDIX B ADI SCHEME FOR MOMENTUM EQUATION The ADI scheme can replace the explicit scheme in the momentum equation in order to eliminate the Courant condition for the time step size.iillaY~i~~l_ ^YII^^~II ( .i. B. B. Since there are two steps involved in solving the momentum equation in each direction. The time step size limitation in the ADI scheme comes from the semiimplicit treatment of the nonlinear convection term. They are formulated for a natural convection problem where the gravity acts in the zdirection.3 are the z and x direction sweeps of the x direction momentum . Equations B. there are six sweeps in a three dimensional problem and four sweeps in a two dimensional problem. The sequence of the sweeps is arbitrary and depends on the problem under consideration.1Eq. The most important sweep should be given the first priority and most updated information.
3) . sweep of the z direction momentum equation is treated as the last one in order to be given the most updated information. Equations B.Rxi+ jui+ jui+ j RX * n+l (B.u Lr * n+1 .2 and B.2) n+l i+j * .P z  n * RZij+ ij+ ij+ (B.2ui+.182 equation.u n+l ui+2 2 n+l Ar2 +2u j n+l ui j . u n * u j i+ Ui+D Dt j + * n i+ j+ Ui+ j .2ii+ Ar 2 + Wilj+ .ui+ j At  n+l * u i+lj i+ .4 are the x and z direction The z direction sweeps of the zdirection momentum equation.Wi+ Az j Ui+ j 1 ui+ j+l * . 2 Az + Ui+ 1 P  n  i+j i+ jui+4j (B.1) * * n * *  ij+ At Wi+1 j+ ij+ + i+Ij+wij+ uij+Wilj+ 1 P n~ . ar .
3. and B.4 have given almost identical results with the explicit scheme for a time step size below the Courant limit.1.4) Equations B.2.+ n+1 n+l Az 2wi + w _P z P0 zP g z RZ n+l ij + wij+ 3wij+ (B. .183 n+l i+ * wij+ * i n+l +lWij+ * n+l zW. B. w At n+l Wi. B.
184 APPENDIX C COMPUTER PROGRAMS The files. Fig. are the computer programs for calculating the numerical values of the analytical solution in the problem geometries. 5.4. There are When there are inflow and outflow. 5. y and z directions. Then four crossflow diffusion constants at every interface are summed up to give a total crossflow diffusion constant .2. the crossflow diffusion constants can be calculated in a simple way as in the first part of the program. In the second part of the program the flow split is considered such that every outflow velocity is partitioned into four components in a three dimensional case in proportion to the velocities of the four neighboring surfaces in contact with the surface under consideration.3(B) and Fig. or all inflows in x. a flow split occurs as shown in Fig. are assumed to be zero. two different cases treated separately.8.. The file. y and z direction mesh interfaces. When there are two outflows in any of the x. 'progl fortran'. 5.3(A).8. the calculational logic of the mesh interface implementation of the corrective scheme in section 4. is the program for testing Fig. All the inflow velocities The component velocities are used to calculate the crossflow diffusion constants for eight subspaces in a three dimensional coordinate space. 'difprg fortran'.3. 'prog2 fortran' and 'prog3 fortran'. 4.
The total crossflow diffusion constant can be calculated in the same way for every mesh interface. has four neighboring subspaces. the positive x direction. A portion of the VARR program is introduced to show how the ADI scheme is implemented in the energy and momentum equations. The subroutine 'tdm' solves tridiagonal matrix problems by forward and backward sweeps. The ADI scheme for the momentum equation is given in Appendix B. It is used to solve a one dimensional implicit finite difference equation in the ADI scheme. .185 because one axis direction.
215.3) SlOP END PRO00210 PRO00220 PROO0230 P1ZOOO240 P1I000250 PROOO260 PROOO270 PROOO200 PROOO290 P11000300 PROO0310 PRO00320 PRO00330 PROOO00340 PRO00350 PROOO360 PROOO370 PROOO300 PROOO390 PROOO400 .COF*DSIN(II*HPI)/(1.EO.Tl.2) PI a 3.26 II IF  IP  I GO TO 55 ALN = DSORT(CSO*II*II*YSO) (II.5F11.ALP II FOIRMAT(F7.14 55 ATOT * THIODEXP(ALN*XO)/2.1415926536 C = U/(2.YO) CCOrrF 2.O) ALNX .2.N FILE: PROG FORrRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT REAL*0 (AH.0 44 YY I IIPI*(UJ0.0 00 20 IPsi.U.5)/NY ADD * ATOTTDCOS(YY) SUM T(d) v SUM + * ADO * DEXP(C*XO) 20 CONTINUE SUM 10 CONTINUE WRITE(G.I=l.0 IHPI DO SUM 10 JJI.NY) 100 FOnMAT(IOX.3F7. C*C XO " NXOX YO * PROOOO0 PROOOO20 PROOO30 PROO0040 PRO000OOOO P14000070 PRO000OGO P11000090 POOOOT100 PROOOO10 PIIOO0120 PRO00130 PROOO140 PROOO150 PRO00160 PROO000170 PRO00100 P1.OALNOEXP(ALN.XO)/(ALNC) GO TO . AAN*2.NX.IALNIC)*DEXP(2.NY.OZ) T(tO) DIMENSION REAO(10.0*TII/PI PI/2.O*ALNXO)/(ALNC) COF * CCOFF/lI AAN .100) (T(I).000190 PI00O200 NY*OX YSO v PI*PI/(YO.4ALNX) ATOT .NY  0.1 ) DX.04ALP) CSO .
DSIN(BEN+Y3) Y4C .0 . TC(IO) TH14.2.0 SUM3X 0.21)*DX = X 4 YO/2.0 = YO/2.0 X = (JJO.1.5) EXAL = OEXP(ALNOX) XAL = DEXP(ALN'X) EXALN D EXP(ALN*XN) XALN = DEXP(ALNXN) IF(II.X = 1.0 SUM2 = 0.0 .0 X + XO/2. NX.5) XALO = DEXP(ALN*XO*O.NY2 0.OTHIH/PI 11PI = PI/2. 11 FORMAT(F7.FO.OALP) CSO = C'C XO * PROOOO O PRO00OO20 P R000030 PROOOO40 PROO00SO POOG0050 PROOOO000 PROOOOOO PROO0090 PIR000 100 P11000110 PRO00 120 PROOO 1. IC1l REALi0 (AII. 1415926536 C .(ALN#C)*DEXP(2.21 II = IP  I ALN = DSORT(CSO + IIII*YSQ) EXALO = DEXP(ALN*XOO.OCOS(OEN*Y4) Y4S DOSIN(3EN'Y4) ADO * AAN*(EXALO + ALNX*XALO)*YOO PROOO520 PR000530 P1OOO00540 PR000550 . TL(10).DCOS(BEN*YI) YI Y22 • DCOS(BENY2) OCOS(BEN*Y3) Y3C Y3S .O) GO TO 55 ALNX . ALP READ(10.5*YO .DCOS(BEN*YC) . GH(10).O*ALN*XO)/(ALNC) COF = CCOFF/II AAN .COF*DSIN(II*HIPI)/(I.0 SjU3 " = 0.0Z) DIMENSION TII(10).p A FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 MPI.215. GL(10).5+YO 00 20 IPs1.11) DX.0 = 0.0  RTO DO =  OSORT(XXX) 10 JJ=1. NY.+ALNX) BEN 11*PI/YO YOO .3.0 SUMaX SUM4 XN vI Y2 Y3 Y4 • 0.XN = XN . U.3F7 .2) PI .0 SUMO SUMI r 0.30 P1OOO 140 POOO00150 900440 PROO0170 PRO00180 PROOO 190 PROO0200 PROOO0210 PROO0220 PROOO230 PRO00O240 PRO00250 PR11000260 PRO00270 P1O00200 PR000290 PROOO300 PROO0310 PR000320 PROOO330 PROO0340 PRO00350 PROO03GO PROO0370 PROOO300 PR000390 P11000400 PROOO4 O PROOO420 PROO0430 PROO440 POO450 PI0OO0460 PR000O470 PROO0480 PR000490 PROOOSOO PR00510 DX'NX YO " DXNY VSO " rPI*P/(YOYO) CCOTF = 2.O.5)*DX = YC = (JJ.0 NY2 XXX NY/2.U/(2.
I WRITE(6. ALNX*XALN)*Y3S ADO3Y * ADD4X * AAN*((C+ALN)*EXALN + ALNX*(CALN)*XALN)*Y4C + ALNX*XALN)#Y4S AIDD4Y * EN*AAN*(EXALN GO TO 77 55 AOOI = TIIH*XAL*O.5 ADO3X 0.00) (GL(I).0 AOOX a 0.NY2) = I.5 A002 * TMIIIXAL0.5 A)DO * TIHl'XALO*O.II.NY2) WRITE(.0 ADOvY * 0.tOO) (TL(l).II.(" % FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 ADOI w AAN*(EXAL+ALNX*XAL)*YII AD02 .NY2) WRITE(6.II.EXALN + ALNX*(CALN)*XALN)*Y3C EN*AAN#(EXALN .5) SUMI*DEXP(C*X) SUM2*DEXP(C'X) (SUM3X+SUM3Y)*DEXP(C*XN)/RTO (SUM4X+SUM4Y).SFII.DEXP(COXN)/RTO PROO000560 PROO0570 PROOOS80 PROOO590 PROOO600 PROO0610 PROOOG0620 PR000G30 PROOO640 PRO0650 PRIOOOG60 PROOG0670 PROOO680 PROOOG90 PROOO700 PROOO7 O PROO0720 PROOO730 PROO00740 PROOO750 PROOO7GO 1PI000770 PROOO700 P000790 PROO000000 PROOOIO1 PIOO000120 PROOO030 PROO0040 PROO000050 PROOO060 PROOO870 PROO000 PROOO090 PRO00900 PROO0910 PROOO920 PRO00930 PR000940 I0760 00 oo I .tOO) (TC(I).AAN*(EXALALNX*XAL)*Y22 ADO3X v AAN*((C+ALN).3) STOP ENO SUMO*DEXP(C*XO0O.0 ADO3Y * 0.100) (GH(I).0 77 SUMI = SUMI + ADO SIIMO a SUMO + A000 SUM2 a SUM2 + A002 SUM3X = SUM3X + A00O3X SUM3Y w SUM3Y + ADD3Y SUM4X s SUM4X + AOO4X SUM4Y a SUM4Y + ADO4Y 20 CONTINUE C TC(dJ) THI(JJ) TL(J) Gi(JJ) = = GL(dJ) 10 CONTINUE C WI ITE(6.NY2) = t.NY2) WRITE(G6. 1OO) (TH(I).I 1OO FORMAT(IOX.
NY.22 II IP I PR00360 PRO00370 PROOO3UO P(1000390 P000400 ALN = DSORT(CSO + II*laIYSO) GO 10 55 Ir(II.OtTIIIM/PI HIPI = PI/2.NX*DEXP(ALNXI))*YI I ADD2AAN*(DEXP(AI.0.0) DO 10 00=1.0 XC=Y0/(2.0 SUM3X = 0.0 NY2 nRo P000O070 13100000 PROOOOOO PROOO0000 PRO00110 = NY/2.0)/NY2 X4XOYO/(2.DCOS(iEN*tY2) YV3C = DCOS(BEN'Y3) Y3S = DSIN(BENtY3) YAC = OCOS(3EN*Y.RT3)+(dJO.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN = COF*OSIN(II'HPI)/(I.5)*YO/(2.4) Y4S = DSIN(6ENlY4) ADDO=AAN*(DEXP(ALN*XC)4ALNX*DEXP(ALN*XC))*YOO ADDIAAN9(DEXP(ALN*XI)+AI.0*RT3)+(J0.0*ALP) CSOQ C*C XO = DX'NX YO = DX*NY YSO = PI*PI/(YO*YO) CCOFF = 2.0*RTJ))/NY2 YIYO/2.S)#(XOYO/(2.5) * (YO/3.0*RT3))/NY2 Y3=(JJO. G1(10).O' XO/RT3)/NY2 XI=(dJ0.O*RT3)+(JJO. I 1) DX. U.3F7.0(JJ0.C.(dJ0.0*RT3#NY2) Y2=YO/2.S)*(XO/RT3YO/6.0) PROO000120 PROO000130 PROO0140 PROOO150 PR000O160 RT3 = C SORT(3.5)*(XO/RT3YO/6.NY2 SUMO = 0.SO5T(2.XO/RT3.N*X2)+ALNX')OEXP(ALN*X2))*Y22 AD003XwAAN((C+ALN)*DEXP(ALN9X3)4ALNX*(CALN)*DEXP(ALNtX3))Y3C PRO000410 PROO420 PR000430 1R'000440 P1100050 PR00O4GO PROO0470 PlOOO400 POO11000490 PROOOSOO 1100005 10 PROOO520 P0000520 P1OO0530 PROOO040 P1000550 .5)*(XOYO/(2. NX.5)*(XOYO/1r3)/NY2 YC=(JO.0 SUM4Y = 0.EO.0 SUMI = 0.2) PI = 3.C P11000220 PROO0230 PROO0240 P1000250 PROOO2GO PROO0270 P11000200 PROO290 PR000300 PROO03 10 P 11000310 PROOO320 00 PRO00330 PR00OO340 PROO0350 00 20 IP=1.O*NY2) .215.02) 111(10).5)*YO/(2.0 .1415926536 C = U/(2.0 SUM3Y = 0. FILE: PROG3 FORIRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT DIMENSION C 11 REALtO (AH.DCOS(UENOYC) YIl DCOS(OEN'YI) Y22 .0. ALP PROOOO10 GL( O).o) ALNX = (ALNwC)tDEXP(2. rORMAT(F7. ( 10).0)/NY2 X2=(JJ0.5)YO/(2.0 PR00O170 PR000100 PROO0190 PR000200 PlO00210 SUM2 = 0.0*NY2) X3YO/(2.0+(dJO.2. TL(IO).+ALNX) OEN = II'PI/YO YOO .*0RT3#NY2) Y4=YO/3.5)YO/(2 . TIIl. TC(1tO) PROO0020 PROO0030 PROO0040 PnOOO050 PROOG000 READ( 10.0 SUM4X = 0.
5+SUM3Y*0.NY2) WRITE(G.NY2) WRITE(6.OEN*AAN(DEXP(ALN*X4)4ALNX*)DEXP(ALN*X4))*Y4S GO TO 77 55 ADDOTIIII*DEXP( ALN*XC)*O. 100) (GI(I).NY2) tOo rORMAT(IOX.0OEXP(CXI) PROO0670 PR000600 PROOOG90 PRO00700 PRO000710 PROO00720 PROO0730 PROO00740 PROO00750 PR 00760 PROO070 PRO00790 PRO00810 PRO00020 P'11000830 P1i000840 PROOO00OSO PRO006O PROO00070 PRO0000 PR000890 PRO00900 PROOO0910 TL(JJ)=SUM20+EXP(CX2) GL(JJ)=(SUM3X*O.5 ADD2TtII#DEXP(ALN*X2)*O.100) (GL(I).SUMI + P1I000620 PROO00630 PRO00640 SUMO w SUMO + SUM2 a SUM2 + SUM3X * SUM3X SUtM3Y = SUM3Y SUM4X SUM4Y * * ADO ADO A002 + ADD3X + ADO3Y SUM4X + ADD4X SUM4Y + AO04Y 20 CONTINUE TC(JJ)=SUMO*OEXP(C*XC) TtI(JJ)=SUMI.I= .11.NY2) WRITE(6.0 ADDOX a 0.I1.3) STOP ENO .5 ADDH11H*DEXP(ALN*XI)*0..100) (TH(I).I=.0 77 SUM1 .5*RT3+SUM4YtO.100) (TC(I).0 ADO3Y = 0.5 ADO3X PROOOSGO PROO00570 PR1000500 PROO000590 PROOOGOO P1100000.NY2) WRITE(6.SF11.FILE: PROG3 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 AO3Y=BEN*AAN*(DEXP(ALN*X3)4ALNXODEXP(ALN*X3))*Y3S ADI)4XAAN*((CALN)*DEXP(ALN.I=I.X4) ALNX*(CALN)*OEXP(ALN*X4))*Y4C ADO4Y.0 ADO4Y = 0. PROOO6tIO .1OO) (TL(I).5)*DEXP(C*X4) 10 CONTINUE WRITE(6.5*RT3)*DEXP(COX3) GHi(JJ)(SUM4X#O. = 0.
DIFII'IS2 rGR RfGTI .0) IF(VP.0.2) UU * UI + U2 VV .UU*DELR*( PPZ + PPT )/PSUM DIFZ * VV*DELZ*( PPX + PPT )/PSUM DIFT .VELZ(0).WW/DELTII PSUM .0 RFGCR 0.LT.OR.O.EO.O.GT.DIFRA(0)1.IFZA(0).O.0 n I FOO 130 OMTI m 0.O 151 4 0 ISS1 O 156 O DIFOOOIO DIFOOO20 DIFOOO30 OIF0040 011:O0050 DIFOOOGO DIF00070 DIFOOOTO DIF00110 RFGRI = 0.4 ~ FILE: O FI'IlG FOITIAN A VM/SP COINVERSATIONAL MONITOR SYSTEM PACE 001 DIMENSION VELX(0).VM.5*( VM V2 0.O.IS .DEL'TII.GT.VELT(O).DELZ.VP.L1.GT.0 IF(UM.5*( WP C IF( ISA.O) IF(WP.EO.OR.0) IF(WM.ISt + IS2 ISO IS3 + IS4 ISC ISS5 + ISG C UI = 0.DIFZ'IS3 * DIFZ*IS4 RFGr GO TO 00 AOS(UM) + ABS(UP) DI F00170 DIFOO IGO DIFOO00190 DIFOOlOO ISI IS2 IS3 IS4 155 ISG .5+( UM U2 0.AOS(VM) + AOS(VP) AUS(WM) + AOS(WP) DIF00320 DIFOO00330 DIF0340 DIF00350 01F00370 D)1F0030 D roo390 DIF00400 DIFOO4 IO 1FOO00420 I) FO0430 DIFOO440 DIFOO450 DIFOO4GO )DIFOO470 D r004flO DIFOO4O DIFOO500 DIFO0510 DIOO00520 ODFOO530 DIFOO540 DIF00550 bMtt .0.EO.LT.5) DELR.( WM W2 0.ISC.0 RFGTI .UM.O 15S3 .0.WP 1SI * 0 152 .0 ) C DIFR .WM.0 OMT .2.EQ.O.5*( UP VI 0.PPX $ PPZ + PPT GO TO 500 IF( PSUM.WW4DELTII#( PPX + PPZ )/PSUM C RiFGII .0) IF(UP.t = 1 = i v 1 * .DIrn*ISI .5.2.) ISA .DIFTA(0) REA( 10O.0 RFOT 2 0.5*( VP WI * 0.VI + V2 WW = WI 4+ W2 PPX .0) IF(VM.UP.O. bt ts .I 01o00200 DIF00210 DIF00220 DIFOO230 DIFOO240 DIF00250 DIF00260 DIFO0270 DIF00200 OlFOO290 DIFOO300 DI F003 10 I'I .UU/DFLR PPZ * VV/OELZ PPT .
0OPF Z) W2 *P FX' IF Z WI'.X(I)'DELR*( PPZ + PPT )0 'PSLJM DIFZA(1) m VELZ(I)*DELz#( rrX +.0PFT..( .PFT) WI#PI'X#PFZ WI'*ix( .0PFX)*PrET V2*( 1.0PFZ) DIFO00650 oDI roofGo oir00670 nir006s0 mrof0069 ot rooioo DIF00710o VEIX(S) VELX(6) VIELX(7) VELX(0) VEL7( I) V El.EQ.0.( 1. Z(2) vrI.0.0 PPX .Z (3 ) VELZ(4) V EL ( 5 ) 011'00720 DIF00730o DI F00740 ui roo75o DI F007GO ui FO0llO DIF 00700 0D F00190 0117OO8OO OIFOOII10 o)1roos~o OI F001130 DI FOO 40 otrooaso 01 roooGo Dl FOOD 70 VELZ( 6) VELZ(7) V EL Z( 0) VrELT( I) VELT(2) VELT( 3) VELT(4) VELT(5) VELT (6) VELT (7) VELT (0) Dl FOODO DIFOOR90 0 fF00900 01 FO009tiO 01F00920 DIFOO930 oDIF 00940 OlE 00950 Di FOO960 Dl F00970 0 F0090 01 F00990 oiro0tooo W2$( t.0PFX)*PFZ DO 30 11.VELZ(I)/DELZ PPT .0.0.VELX( I)/I)ELR PPZ .o 'FX'( 1.0PET) Vt .NE.0 oiETA(I) .0PFX)o( t.NIE.0PFT) U I PFZ (I 1.OPFX)O( I.PFX*PFT V2* I'F X'PF T V2.0.0 Ir (W1#W2).0PFT) U2*PFZ*( 1.FILE: DIFPRC'.PrT )/PSUM DIFTA(I) .PFT u i. oi rooG to Dl F00G20 D1F00630 01F00640 Ut 'PFZ .O.0PFT) V t.0 30 CONTINUE RFGRI RFGR RFCTI  DIFO 10 tO 011:'01020 DIFO 1030 01 FOt040 DO 1050s DI FOtOGO DIFO 1070 01000 DIFO 01 FOtO0J0 OIFOi tOO DIFRA(5) v DIFfIA(t) aDIFZA(2) + DIFRA(6) + DIFRA(2) + OIFZA(3) DIFPA(7) + OIFRA(3) + DIFZA(6) + + + OIFRA(fl) OlfRA(4) + DIFZA(7) r % I .0PFZ)*PFT U20( 1.( i.V(LT(I).o VELX( 1) VCL~X(7) VELX(3) VCLX(4 ) PFX i'rz PFT   U1/(UI'U2) Vt/(VI14V2) Wl/(W14W2) r DoIooseo 01F00590 DirooGoo.0PFZ)'$( . ronTRAN A FILE: CON4VERSATIONAL MONITOR SYSTEM DFPRG AVM/SI' FOTRAN DI F00560 DIF00570 PACE 002 PG 0 OMT GO a 0IFT'156 TO 999 00 PFX PF Z *0.NC.0PET) Vl*( 1.VEU.IPFZ)s( 1.0PFz) W2 &1EX'&( 1.0PFZ) W20( 1.VELT(I)/OELTH PSUM a PPX + PVZ + PPT IE( PSUM.*( PPX + PPZ )/PSUM GO TO 30 600 OIFRA(I) .0PFX)#PFT V2( 1.09'FX)#(t1.0PFX)'PFZ Wt'(I.0 0.0PFX)#(t.0 Pr T Ir (UI+U2).0.0 (Vl*V2).0PET) Vi*PFX*( 1.0 ) GO TO 600 DIFRA(I) .oPF7Z)'PFT UIs .0 oirZA(l) .) U2 4PF Z PET 1J2'( 1.DELTH.
RFGR.RFGT.RrGT I.2) 7 FORMAT( ItII .6F5.I. FILE: OIFPRG FOR TRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 003 RFGT OMrt OMT = DIFZA(1) * DIFZA(4) a DIFTA(3) + OIFTA(4) DIFTA(1) + DIFTA(2) DIFZA(S) + DIFZA(0) 4 OIFA(D) 4 DIFTA(5) I DIFTA(6) DIFTA(7) D 999 WIllTE(GG) RFGRI .511S ORRY) 200 STOP ENO DIFOi IfO DIFOI 120 DIFOI 130 oDIFO140 DIFO 150 DIFOI 160 DIFO 170 DIOt t0100oo oIro 190 DIFO1200 DIF01220 I I .2) G FOIMAT( III .OMT I.20X.2X.OMT GO TO 200 500 WRITE(G.7) 5 FORMAT(9F5.
P(400).O(N).C(N).NMI .NMI 1+1 )/0( I) P( 1+I) A( 141)#*C( I) P( N(I+ I) 0(111) 3 X(N) *X(N)/0(N) DO 3 1I.0.C .o(N).XN) DIMENSION A(N) .1O9:190 VA1109"IO() VAn0094 t VA11O0d1O '4A110943?0 VAIIW14 40 VAfl0O450 VAI*0946) VAI?94 70) VAR0O9400 VAR094 90 VAIR095C VAflfO!'j 10 VA110IJ520 .D.0t4O01 X( I ) I~) 13( ) 0(i1) NM I *N I 00) 2 1I.I KN a (X(I() X(K) RE TURN END  X(K41)*COK))/O(K) VAR09Z350 VAflo9:160 VA1109370 VA11093001 VAI.X(N).SUarIouT INE TDM(A .
) G0 TOrOOiC D'.L.PM.2D LUGRDX/(AUGOX +~DG'RD)' .F2(J.E230 .KZ2))) .UG.FFX ( .]K .. D2).E0.KS 2.LE0..K) * 0. I vtR I5S0 rooOT 00G2 UC 'S(U(]K) ) VLRi 1210 vAR1E230 VAR E20 U(3KXX) ) LUGc .S(0.K) a LUGDX.P.C "hT..ItD.0 OOC COI~'T2]'UE C VAR P182 vLR a210 VAR E220 VAR.0 D'rF2( I .DFFUSI'O =.0.EO.K) z 0.S( 4 .:X GO TO 40000 "0OO D.I.E0.KBP2 " (3. 1) GO TO 400 IF(K.UG. iB10 . C C C C U AND v LDE VELOCITY EDU.FX(3.1) GO T0 49998 SOLUTiON FOR SIE.LT. I O50 VI8090 K1 K12 2 KEP CONSTA'NT CLLCULATION C CRO5SFLD'.I.C .G  V RIS150 VAR.E1O 0005 FP" * IF(.L!S( (]KZZ) ) .)K2NC " (Ki)'NWPC . DNF..3.xx'..K t KXX a jK .TIONS SECT ION AD !ST ]F(ICLL].LT.VGRD2/(..5(V( K)*:(.P2) GO TO 40002 (U(iK)*U(KX))) LUG = LS(O.NPC o1 ]F(K.O.K2NC JKZ2 .G.4 SwEEP IN XDORECTION !KS'= a DORKS VAR17980 VAR 7 990 • VAR i8000 V/R 160O0 VAR 8010 VA 18020 VAR1C030 VR180^0 vR 14BO1O VAR.K) • AVG.CR.EO. 1CONV.JSP2 DO O40000 DO 4000 K.S . DA'' G.
3F(3~. .I) K2WC' + II)K NC *K2NC "00Z0100 K=2. ( '(K ULH C CON'.2)GO a TO 29990VALR iS.I DIFFCD(3KL2) D!~FFCD(IKL3) *0(!.]BR ( '.ET/TADD3 VAR IS420 VLO IS440 3F~lDw.R2CLI.S(UftH)(UCUQ)  VL21S660. 054 (K.1. a uJD(2K) )0.0. UD( 3K) UL UR UT Uo PK) UD( 3KP) U0( 3KMS) DCR DCL D!FFCD(IKL) D) F CO( KL.1 SRKSrI a * N BUOJY a LP0E7.E003 DO E0100O ) K GO TO50205 222. VLRIE6t0 V4RI15650 VAR 15690 ULWM'(uLLIC) ( IR'(L'C4UV) VARIE *7 DO.5 UODUMKS) 16D(1PK) * .C RD20 3KS0 S3G'~ to VAR 8340: VARIS5360' RDZDRD2 VAR i !:0..5 PC) ')...0.E.KSPI I KL I lK z lK 1A 0.I ) I KIPC # (KI)'tN'PCL VAR16D VAR 11510 VARi18520 !PK a 31.P( TJ"'.PKKNUO(IPK) ). KS I KLS a3)K I KP a IK  NP 2 I N1PC VAR 185501 YLR 15560 VAR 1C570 VARias590 VAR 86201 VARi1Es30.5 )r0.'O('.
OSZ6 M7A OdI'N Al dN O7I~t 17A OdM ( A) ONZ Nt + 1IOZN4(Vt) * + 3NZ)(11) + t bi > >4 E A c TA .o.t) 2 td9N'Zt oo0to oa 3flN tLNOD s0?_O3 M 6 dVA + NAcns + ZntO * (Nt)d(dXt)d ).( (nf1l)(Himlss7 (In3n)(uLM)S9V ) ZNQ.uQ + + t (Nt)0ftl x = XIt)f ( oo..(.) .(?nOn) i13( fl2f) unc.onH.)10 6666a 01 0 MfN i NGO 07 6 "VA ZNOO NCO ZOM( + (Xt)0M £ ) O~t6tVA (~ ~ ..% At t x NdE ONCA + lA b + A t + OZ~N(Lt) OdAN 00gla aVA 06!.(iim1 (.Cw'))'A.(onin) M0'a(On. (W (4I) 09O6t~u7A CA ~g0 *0 H IL~ Iun S:)o ( (dN I)an +(>(!M)Onl 'S0 06SLIA OEOG bVA 0L6O t'7A 0CS8~ 100 VA )O (dm)OoA1 0669'.=OStd7A 0506LdV  ~ ~ ~ ~ Hm )CC7r~! Sall*~ ZN07J ±.o2m.vt Z1t0 7 Z(J~Si0(o(nsn)(a)a (1n.'N(tX) + 06Md~7A 1 Et a 6Nr NZ At 2 SW SWA r r ). Qti 0*0 o3 a itflt NACAS dsm~rsN td3t*Z9t Oozos 00 00ZO9 00 0LL8Z!VA 0SL8dVA (Zito ZNOO YNOO0 YO + ) 00L3'ttVA 06L9tMVA 0 LStMVA (At)d(Adt)d  11 )*..7A 0060tMIA 0L6Sgv7A 0968LVA I 3a CA~ SimI) 0L96 t'7A O9S63tM7A j0 0 VA d061Nt) AoflsaNx0As + Od.+fl)IdM Y(Qd.8~ VA 0899tMVA (oaQN 3NC647.
IMKS IKt.ABS(ULT) ) .SIGMhARDZDODCB E1(J) = RDT * 0.$ ]K .]K + K2NC IKP a 1MKS 1K + NWPC ]K K K2NC NWPC 1KIAS = C DCR .5 AI(J) a 0.ABS(VRT) ) .'5 WRS (W(1KMS) * W(IPKNWPC))O.5RDZ*( VRT * ASS(wRT) .5( U()]KS) * U(IMKS+NWPC) ) Ai(J) = 0.IBR DO 81100 K2.KBR) C DO 82100 Ka2.16PI 1K a 1 * (]i)K2NC + (K1)NWPC IX x (K2)ISR + I 1 E2100 W(]K) * x(Ix) 00 823000 K2.DI.IKBR) CALL TDM(A 1.RDX.5*RD2( WRT .C1 .8e1.X..5RD02( WRe . C 00 l8100 I=2.KER D00 82000 12.ULT * ABS(ULT) 81(J) .5( U(IK) + U(]KP) ) ULT C a 0.DIFFCO(IKL) DCL a D0FFCD(IKL+2) URT x 0.( DCL * DCR ) C1(J) z 0.KBR + K 81100 U(IK) a X(IX) C DO 82000 K=2.1SP J a (K2)]IeR * 1 .X.AeS(WRB)) .ROT * 0.KSPI ]K = 1 + (11)K2NC + (KI)NWPC 1X a (12).SIGIArRDZD3DCT 01(J) a U(IK)ROT 81000 CONTINUE C .SIGt.NWPC DCT 3 DIFFCO(IKL+1) DC.VRE + ABS(WRB) 1 * SIGMA*RDZD.C1.KBR DO 62100 1=2.5*RDX( ULT .5RDX( URT 1 + SIGMARDXD. VAR19310 VAR19320 VAR19330 VLR193AO VAR19350 VAR19360 VAR1S370 VARie380 IKL a C 1 + (1)'K2NCL + (K1)NWPCL IPK .1 ]K £ + (1])K2NC + (Ki)NWPC ) VAR 19260 VAR 19270 VAR19 280 vLRlg2O VAR 19290 VAR19300.0.Ke1P ) VAR 19390 VAR19400 VAR 194 10 : VAR19420 VAR19430 VAR19440 VAR19450 VAR19460 VAR 19470 VAR19480 VAR19490 VAR19500 VR 19510 VAR19520 VAR 9530 VAR 19540 VAR19550 VAR19560 VAR19570 VAR19580 VARi9590 VAR19600 VAR19610 VAR19620 VAR19630 VAR19640 VAR 19650 VAR19660 VAR19670 VA2 1960O VAR 19690 VLR 19700 VAR19710O VAR19720 VARIC720 VAR197540 VARS19760 VR 197 70 VAR19780 VR 19790 VAR19800 .]K .5. = DIFFCD(IKL+3) WRT 3 (W(3K) + W(iPK))O.W(1K)RDT 82000 CONTINUE C CALL TDM(A .SIGMARDXD*DCL ACBS(URT) .( DCT + DCS ) C1(J) a 0.'RDXDDCR D01(J) .K2NC .( URT EAS(URT) ) .
40 C0~7A O CO0d7A NAoC1s (A(I)a .)ssv . )'ZQdS50 )Zc!S*  a ' a  (HS.v4)tS ~m * io1.3nwNtJ.fl.L 0.Htin 0~Q0~*/'.~) Al NP 1=I000ca 00 . 3)3QO.ycajSpo Hlfl)bQS 00L661 dVA 096tdV'A C(S~wt)fl + (Ar4)fl )s0o )'s0o Hlfl C(xid[)n + own)f  Htf 0OS66taVA * 0bVA I OS6J&1VA Z+1~~)41 2jE 1d)a O1'66t7A 01L61 J7A 3dI% () UMNN () >1I a SWN1 * D ).ZC(3aE1~S (d~t)d  00oo' ((HIiA)SS7 I )'ZQZ .~tSC(Hifl)ss b3(3 ).3 00 000O~d7A "L661 aVA (AI  (Ad 1) d !OSGOZi VA OG61 !VA 03661 SVA b33aXac~tjS ( 100 (Hll)SV H~ln   C(Hzn)sev  Ya   ~ A +Laa Hfl )xCi~s~o = x (r)La ()9 t ()1 2 (H. z (r 01)~S~7~).(1) L >1 * (~~t A4d! zz IN I s)Zt! + Afl9nszAofl3 (SCOP 001 0OZ17A 060ZZ2A OSOOZZ17~JA 1 OdIA'Na)) *  AC470103t1)1E 0*0 a NAC(3 a9AzsA 007 OC Idgl*Zal OOOVS 00 OLOOZ!7A 0900ZM7A * ()t(z>q) :)NN41) + I a as9tz:t 001C3 + Cxt)y 2 (x)Xfl( ( 00t ES A4t cc tdaA'Zx4 001..LG!1CE)PA HIM.' ).v%4!)iS +.'%)SS'7 + HSA .Q ( (S'feXtiM 094=J7A OSZO0t7A (>I).(HL')S.0xCd. ).mNtAz + + ONA(lt.S'Q g* 2 S = CI)' 1 C + ># O oc OZMVA ~dN OdMN )Z' 1I A W> = S AVI + A!t d>'[ 06t0MIA M17A 0910~1VA 0C10aVA +N *) 1:)dfMN'()) + IONZ)A(Ll) OdAIY( .
5RRC*RDX CFC .SIE(IPK) S*EL .HDX RRC a I.0P.U(JMKS) vW = W(IKMS) AUC .B.NWPC CFR = CF(1PK) CFL .S) SIECO .I) GO TO 55000 DCR .1) GO TO 2040 ISEO = ISEO + 1 IF(ISEO.1CALI.O.S]EO(]K) TSRA TSLA . VAR20420.5(TS(IK)+TS(IPK)) . IBR(K2) + I ' (1i)K2NC IK = i + (K1)NWPC IKL a 1 + (11)K2NCL + (K1)NWPCL RC a FLOAT(I1)bDX .1 (12)(KER1) 84100 W(JK) a X(lX) C 89999 CONTINUE IF(ICONV.CALL TDM(A .CF(]MKS) UC c U(IK) WC = W(!K) UL .X.D1.DASS(UC) AUL .2.EO.12 J .NSEO) 33333 CONTINUE 15E0 a 0 DO 60000 K=Ki.EO.K2NC ]K IK 7]KS IKMS = IK .DLeS(UL) ALC = DAES(WC) 1wS * DASS(WE) = SIE(1K) S!EC SIER .NE.CF(UK) IF (CFC.1SP1 00 84100 K2.25*RRC CYRX = CYLO0.5(TS(]K)+TS(lMKS)) VAR203G0 VAR20370 VAR20380 VAR20390 VAR20400 VAR20410 . VAR20430 VAR20440 VAR20450 VAR204GO0 VAR20470 VAR20480 VAR20490 VAR20500 VAR20510 VAR20520 VAR20530 VAR20540 VAR20550 VLR2050O VAR20570 VAR20580 VAR20590 VAR20600 VAR20610 VAR20620 VAR20630 VAR20640 VAR20650 VAR20660 VAR20670 VAR2060 VAR20690 VAR20700 i VAR20710 VAR20720 VAR20730 VAR20740 VAR20750 VAR20760 VAR20770 VAR20780 VAR20790 VAR20800 VAR208 10 VAR20820 VAR20830 VAR20840 VAR20850 VAR20860 VAR20870 VAR20880 VAR20890 VAP20S00 .LE.IKBR) DO 84100 l=2.K2 DO 60000 13a1.0.SIE(IMKS) SIET .0/RC OR = CYL0.Ci.DIFFCC(IKL) DCT .KSR IK IX a I + (I1)*K2NC + (Ki)*NWPC + K .SIE(IKP) S]EB * SIE(]KW.DIFFCO(1KL*i) DCL a D!FFCO(1KL+2) DCB * D]FFCO(JKL*3) ]PK )KP a a + K2NC * NWPC = 1K .
.OSrlth7A OV7 t!VA t. 10 + ( 0S669 01 00 (V03VIJfl S0. (nN*31.1! Al z(IJ)L3 (NOfl'Ol V(ISJX017) Ot)t0dVA 0001ZlA N7 dlxiol ((L1)x~JNG4(IV~~sNQ)ts0 .NOD (L03e. 06C Zd7IA ZltrEzf 00 OOOL CO 0SCtLb7A NOI0t1D3JCZ NE d33MS CNZ 0L~tZZ7A LCECE OZ :tZ7A 1 05 3 3lNTI. C0 r~t (r)LC 00~ttIdVA 060k ~wVA 0OLt dVA 00009 01 13t5. t7) .x)j! 2(AI1)3rS oo0t 9 06CI ~1VA' 09ZI dVA 0LI."ai 0*0 000 09ML VA 3flNIINOD 00009 =(fil) 0O~VA 0Lt t aVA OOtZVA 06tt~7A XC(QXC + dAO)dQZJlj + 01 0 00009 02.I vo 0009 0:9>4 I )c t a.C*~ t )4* .~t J '~1f ((t0AX'1ZJANO ) C dt0* ) * N ) S 4~ I .± 4:7) 21J3 1u L7S1L'q'l OS 60ZMVA (Slx () I s I)SO I ((diE)l+ m1)l)'c = vzs I VIL .(N t0 *090~bVA 0O0I bVA 0O'0t MVA OSOI aVA 0OP0tZSVA (Q QtlY.Q ~ 000 fl co(3 S6 .L.(r'.OlI)03(r v+n .+c a 21J10 to. N *:dM o zi Z)q4s)4 coo 0. t :) a .At~VA OSZI~7A clt'L fa ZAVtAZ'4 0=0i9 00 0=O9 00 I .
1) GO TO 65500 D1(J) * 01(J) .DIFFCO(IKL) = D]IFFCO(IKL+1) = DIFFCO(]KL+2) .5(DNFF2(3.M * * IF(TS(]K).0 70000 CONTINUE C CALL TDM.5(TS(]K)+TS(3KP)) TSeL .21730 VAR21740 A2(J) = B2(J) = VAR21750 VAR21760 VAR21770 VAR21780 VAR21790 VAR2800 VAR21810 VAR21820 VAR21830 VAR21840 VAR21850 VAR21860 VAR21870 VAR21880 VAR21890 VAR21900 VAR2190 VAR21920 VAR21030 VAR2190 VAR21950 VAR21960 VAZR21970 VAR21980 VAIR21990 VAR22000 .01.IF C DCR OCT DCL DC8 C ( CFC.ND.12 DO 71000 KKI.NE.CO(IK)0.IKBR) C DO 71000 ]=m1.5(DNFFZ(I.1) GO TO e9950 Dl(J) = 01() .0 C2(J) = 0.C2.I DI(J) a TGA.0 69950 IF(CFT.CF(IKt S) WS W(IKMS) WC a W(IK) VAR21570 VAR21580 VAR21590 VAR21600 VAR21610 AWB a DLeS(we) AWC DOASS(WC) TSTA a 0.0 DI(J) D(J) + DIFFTDCT*RDZ GO 70 70000 65000 01( ) = SIE(]K) L2(j) a 0.K)*DNFF2( .5"(ws*Aw )'RD2 .TTSPLL .E0.i) GO TO 70000 IF(CFT.K2 Ix = KS.A2(J)SIE(3KMS) A2(J) • 0.0 GO TO 70000 65500 C2(j) E82(J) .5(rS(lK)+TS(lKMS)) C GAT DIFFT DIFFB C C * RDTSIE(]K) .NWPC C CFT = CF(IKP) CFB . 1 * 0.EO.I) )XMX GAMTTSTA VAR21620 VAR21630 VAR21640 VAR21650 VAR216GO0 VAR21670 VAR21680 VAR21690 VAR2I700 VAR217 tO VAR21720 VA.11..B2.DIFFEDCRD2D RDT * (DIFFTDCT+D)FFS*DC6)*A02D .5.NNOPT.RDZ C2(J) = 0.(WC*AwCWS*iWB).X GtA.0.k)+DNFFZ(1.1) GO TO 65000 .0 E2(J) .5(WCWC).DIFFCO(IKL+3) 1PK a 1K + K2NC IKP a !K + NWPC IMKS a 1K .K.NU) GAMT * RPRAN ..LE.RDZ .X.1.(A2. ..5 0.K+1))X.OlFFTDCTRDZD IF(CFB EO.K2NC IKWS VAR21460 VAR21470 VAR21480 VA021490 VAR21500 VAR215 O VAR21520 VAR2rS30 VLR21540 VAR21550 VAR21560 = ]K .0.C2(J)SIE(IKP) C2(J) = 0.(12) + K .0.DIFFTDCTRDZD C2(J) a 0.EO.
09 KzK1.SIEO(IK) KIRK a i GO TO 33334 33337 CONTINUE DO 33338 1=1112 DO 33338 K=KI.K2 IK v 1 + (1I1)K2NC + (K1)*NWPC STOR(I.K2 1K 1 + (]i)'K2NC + (KI)*NWPC 33338 SIE(K) = 0.K2 LWPC=LWPC+NWPC AT TIMEN ) . TRANSFERS VELOCITIES TO STORAGE ARRAY( 2040 KI1 K2xKBP2 LWPC=I .NE.NWPC DO 2.IK r 1 + (II)*K2NC + (K1).11.S)=W(KK+) TO(KMS)=TO(KK+1) VAR22470 VAR22480 VAR22490 VLR22500 VAR22510 VAR 22520 VAR22530 VAR225ZO VAR22550 .5'( SIE(IK) + STOR(I.O) 00 33335 I11i.E0.1) GO TO 33333 GO TO 33339 IF(ISEQ. IK=LWPC IKS212K2 4 1K SIE(IKS)=SIE(IK) U(IKS)=U(1K) W(IKS)uW( K) TO( IKS)TO( IK) TS(IKS)rTS( K) CONTINUE 12]BPI K1=2 K2rKSP2 KK=0 KKL O0 DO 29F9 l.12 DO 33335 K=KI.K) = SIE(IK) 333335 SIE(IK) .NwPC 71000 SIE(]K) = X(IX) C IF(KIRK.K) ) KIRK a 0 33329 CONTINUE C C NOTE. 12 KK=KK4K2NC KKL VAR22010 VAR22020 VAR22030 VAR22040 VAR2 2050 VAR22060 VAR?22070 VAR22080 VAR220s0O VAR22100 VAR22 110 VAR22120 VAR22130 VAR?22140 VAR22150 vLR22160 VAR22170 VAR22180 VAR22190 VAR22200 VAR222A0 VAR22220 VAR22230 VtR22240 VAR22250 VAR22260 VAR22270 VAR22280 VAR22290 VAR22300 VAR22310 VAR22320 VAR22330 VAR22340 2109 VAR22350 VAR22360 V R22370 VAR22380 VAR22390 VAR22A00 VAR22410 = KKL + K2NCL VAR22420 VAR22430 VLR2240 VAR22450 LWPCL = LWPC=I 1 VAR22460 SIE( 1)=S [(KMS) U(1)=U(!KMS) w( ) (IKVS) 0(l)TO(IKMS) TS(i)=TS(IKMS) SIE(]K S)=SIE(KK4+ ) U(KtMS)U(KK+1) W(IKt.
7.APPENDIX D TYPICAL INPUT FILES The file. The files. 'input mom'. .8. Fig. The file. 5. 5.3(B). Fig. results are in Fig. The output results are in Fig.6 and Fig. 5. 'inputd HUH' and 'inputd DM'. 5. 5. 4. The output The files. 'inputn dat'.2. are for comparing Huh's And De Vahl Davis and Mallinson's correction formulas in the problem geometry. 'inputd 3' and 'inputd 7'.1 and Fig. is for calculating the natural convection flow field in a Cartesian closed compartment. is the input to test the stability of the ADI scheme for momentum equation. are for calculating the numerical solution by donor cell scheme in the problem geometry.
0 0.0 1.0 1.0 0.0 0.7 0.4 5 1 1.0 0.0 0.0 0.0 1.0 .0 0.0 1.0 0.0 1.0 0.0 0.0 60.0 0.0 60.013 0.5 6.0 2 5 10.0 1.t20 1.0 10.0 I 3 I 16G 1.4 0.0 0.0 0.0 0.0 60.0 0.5 0.0 0 0 1.0 0.0 0.rTh FILE: INPUT MOM VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 4 4 O STABILITY TEST FOR MOMENTUM 2.09375 0.0 10.0 1.75 10.120 1.0 0.44 0.0 0.0 1.0 1.9 0.2 10.0 0.4 0.0 0.0 2.01 0 1.0 1.0 1.0 0.0 0 0.0 0.0 10.0 0.045 0.0 0.0 1.0 0 2 1 3 O0 2 2 O O ADI 10.0 4 0.4 1.0 4.0 0.0 0.0 0.01 0.0 2 l.0 0.
90 0. 20 1.01 1.5 5 1.0 .0 NATURAL CONVECTION IN A CARTESIAN CLOSED COMPARTMENT 1.2 1.0 68.0 0.20 2 1.7 1.FILE: INPUTN DAT A VM/SP CONVERSATIONAL MONITOR SYSTEM 'PAGE 001 5 10 0 1 0.0 6 6 50.20 030 1.20 0.20 1.0 1.0 1.01 0.0 1.0 1.0 10.0 8 2 I 3 1 3 0 3 0 1 1.0 I 43 1.0 0 0 2 1.0 0.0 1.0 1.0 0.5 1.5 0.0 1.751.0 4.0 0.0 0.20 II I 1.0 1.+20 60.0 60.67 1 1I 2.0 32.44 0.21795 1.0 2.0 0.045 0.20 1.0 0.0 0.0 0 1.0 0.0 677.0 0.4 0 0 0 0 O 0 0 6 2 0.0 0.0 1.
0 0.0 1.0 10.204 0.07107 7.0 0.0 0.0 0.0 1.000 1.0 0.+20 I 0.207 0.0 0.457 1.0 2 1t 1 113 5.003 7 1.0 7.0 0.0 0 0 NUMERICAL blFFUSION .O 2 1.0 1.0 0.0 .0 9.497 0.0 0.001 0 1.948 1.0 0.0 0.II I 11 2 2 1.70711 0.0 1.0 1 .0 0.0 0.511 1 1.0 0.000 10 1.07107 0.0 1.0 0.000 0 0 0 O .0 12 11 2 121 t 5.CASE 3 2.O.063 0.000 0.0 1.09375 0.0 1.01 0.0 0.0 9.0 2 4 2 4 1.0 1.497 2.5 1.0 1.0 0.751 2 1.0 1.0 7.05.0 GO. 020 2.70711 1.07107 111 3 1 1 2 5.0 1.0 0.0 60.0 4 I 36 1.0 0.991 6 1.07107 12 12 Ill 2 5.0 5.0 9.0 1.045 1.000 9 1.0 7.0 0.47.75 0.999 1.063 ( ).0 9.409 1.0 9.9 0.204 0.000024 0.0 0.0 0.911 3 4 1.0 1.01 0 0.OGI 2.5 1.002 0.0 0 2 5 3 0 7 6 1.997 7 1.974 5 1.3 1 I 1.457 1.000 10 1.0 0.0 0.940 1.44 4.0 1.0 1.0 0.0 1.014 0.009 G 1.0 0.0 5.061 1 0.07107 0.026 5 1.FILE: INPUTD 3 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 10 10 0 0.0 0.0 9.000024 60.0 1.014 ).002 0.0 0.0 2 1.0 7.249 3 1.07107 0.0 1.0 0.0 5.0 0.0 7.0 9 10.0 8 10.0 0.0 0.009 4 1.0 0.0 0.02 1.07107 1.0 0.0 0.013 2.000 0 0 O O I.207 0.0 1.0 0.0 0.000 0.5 0.
0 7.0 0.457 2.143 0.0 1.003 0.167 1.0 0.420 1.0 0.0 0.CASE 7 0.07107 711 8 8 2 5.0 0.0 0.0 5.0 0.014 0.911 9.071107 0.000 0 0 0 1.0 1.0 4 I 36 1.0 1.0 0.0 1.045 1.+20 2.626 1.013 2.0 1 9.0 7.75 2.47.000 6 1.0 0.333327 1.0 1.990 5 1.0 1.0 2 7 8 5.0 0.333327 60.0 0.0 0.002 5 1.0 0.633 1.09375 .0 0.0 0.0 7.0 60.0 NUMER ICAL DIFFUSION .0 10.0 0.0 1.0 0.0 1.0 .0 60.0 2 0.0 0.0 0.009 1.0 0.17051 0.013 4 1.0 0.0 3.9 0.0 1.003 17051 1.4 1.4 8 2 1 3 0 0.0 0.0 I 1.00.0 4 2 4 1.01 2.0 1.000 6 1.907 3 1.0 5.0 1.07107 0.01 0.0 2 2 0.0 9.0 1.0 7.0 0.0 0.001 0.0 0.0 2 9.0 5.0 10.01 0 0.00.0.000 0 O 0 0 0 7 1 2 7 2 5.143 0.0 1.44 4.0 0.0 0.0 3 0.07 07 0.0 0.07107 7.0 7.457 2.001 0.0 1.5 1.0 0.FILE: INPUTO 7 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 6 0.0 4 1.0 0.07107 1 1 713 2 5.014 0.5 1.0 1.633 1.3 1 1 1.07107 I 113 7 2 1.0 0.0 1.0 0 0 2 2 1 0.0 1.374 1.
0 1'0l '0 011 '0 sCl '0 LSI *0 6LI 0 G0Z 0 cpz *0 01 01 0 01 oz 61 LI 01 01 91 cI zI 01 6 0 Ia z 0 6ZC *0 OLC 0 &Pt'v O CZi.0 GLO10 ZU10 SL9*0 109.0 CGO.0 0*1 i 0*0 099C00'LOOo 0*0 01 CI0'0 0*0 010 £LC6OO0 0*1 VL9'C) P99*0 OWO' LLC'O 1'CC*0 16L' 9L1 0 100 ~W31.09 0. 0 LG0 .1bl b 690*6 £00*0 LC9 0 90c.1 1' 0*0 0 *1 D99C00 0'1 0IO' 0*0 0*0 0*0 0'0 0 *I0*0 9c 1 t.*I riI 00 0*0 0*0 0i'00 0 10*0 6 v 0*0 0'0 top' .0 01l 0'00' C0'I C..0 0.L Z01 'L L'09C91C P09c91''? 1'Q9C9p 0'19C91' 1 iLO'O 1'LO*0 GL00O LLO'0 £00. 1'0 0. *0 PI 9.0 9?LL '0 G'.SAS 3~~~)Vd kOIINOW 0 0'0 0*0 0.NOIS(fJJIG IV:)IUIW(IN too 39vd IVNOIIVSU3ANIX) dS/W~A v VIIi III)II OIMINI O~N :311A II .09 W '09 0. 6IP'0 O1'1 991'0 LOCO0 9LV'O 060*0 Cc£00 910'0 1 0 C Iz 0 0 0*1 OL+*t Ylz OZ4*1 zlo' IfIAH .u *0 Ul' I 6P I 0011* I 0.0 0 *I O'l 0' *1II1 01l 0*1 0 1 0'L 01 O'1 0.0 09G"0 9Cs£0 661''0 6sv1'.z 10'o 0 1 0'l GL'O 1O1 '0 I 0*0 01l 0'l L910* 199*0 099.
FILE: INPUTO IIU4U VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 2 0.20 5.20 10 1.6603 0.20 1011 1.20 1 1.20 I13 1.0 2 25 5.20 911 1.0 0 2 1.0 8.0 26 26 5.0 .20 2 1.0 0.20 9 1 1.20 2 1.0 I 1 5.0 5.20 913 1.0 5.0 2 25 25 5.0 0.6603 0.6603 0.0 0.
0 0.0 1.0 1.420 0.0 1.5 1.0 1.0 0.0036G0 0.4 4 1 36 1.700 .044 9.9 0.0 1.0 0.463704 9.0 1.0 1.4 0.0 0.246 0.0 1.522 0.75 1.0 1.403604 2.33G 6 1.223 3 1.GOG 2 1.526 3 4 5 G 2.463G04 2.009 0.0 1.0 1.133 10.0 1.OGI 0.0 I 1.0 0.360 0.09375 0.242 7 1.000 0.566 0.0 0.0 0.003660 60.072 ' II 12 13 14 15 16 17 10 19 20 21 22 23 24 0 0 0 I '. 4G1IG04 7.0 0.0 5 1.0 0.0267 0.002 0.920 0.636 0.0411 0.0 0.002 2.011" 0.0 0.021 0.OM 0.0 0.4GI6 0.0 1.016 0.5 1.932 1.0 0.0 2 4 2 4 1.496 0.0 0.0 1.0 0 2 1 3 0 7 6 I 0.rILE: INPUIO OM A VM/SP CONIVEIlSATIONAL MONI(fro SYSTEM PAGE 001 24 f( 0 I 0.02 2.460 4 1.01 0.903 0.0 1.003 0.002 0.20 1.030 0.144 1.0.463604 2.767 0.051 0.0 1.0 0.0 1.0 2 2 2 0.0 60.0 5.076 0.007 0..0 1.0 0.969 0.0 0.0 0.120 2.0 0.260 0.705 0.0 0.949 0.0 3 10 1.567 9.2406 1.156 0.0 0.96G 0.0 0.02.0 0 0 NUMEtICAL IrFFUSION ..0 1.0 0.093 0.5 1.0 1.44 4.0 5.0 0.005 0.003 0.0 0.0 0.0 1.174 1..0 0.0 1.045 1.365 9.034 0.197 0. 1.030 0.0 0.0 0.029 0.026 0.0 1.02.0 1.01 0 0.099 0.463604 2.0 1.0 1.0 0.115 0.302 0.013 2.3 1.0 0.30i9 1 0.0 1.951 0.0 0.003 0.0 60.0 1.
0 .0 0.0 2 5.0 26 2 1.0 26 5.20 1011 1.20 I 2 1.20 5.0 0 29 25 I 25 2 1.0 I*  .20 913 1.10 OM A VM/SP CONVERSATIONAL MONITOnr SYSTEM PAGE 002 2 0. & .20 911 I.0 2 5.20 113 1. GG03 0.6603 0.: INP.SILr.0 0.0 5.0 5. 6603 0.20 1.0 I 5.0 0.20 9 1 1.