This action might not be possible to undo. Are you sure you want to continue?
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
2 Comparison of Explicit.1 Molecular Diffusion 3.2 Physical Constraint 4.1 Mesh Point Implementation Mesh Interface Implementation 4.1.3 Low Reynolds Number Flow CHAPTER 4: NUMERICAL DIFFUSION Truncation Error Diffusion 4.2 Limitations 2.3 Tensor Viscosity Method Corrective Scheme by Huh 4.4.2.4TABLE 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 2.3.1 4.4 4.3 Discussion of Skew Differencing and Corrective Schemes Raithby's Scheme 4.1 Assumptions 2.3.1 Problem Statement 2.2 ke Model 3.1 Derivation of Turbulence Equations 3.4.4 Solution Scheme 2.1 Continuity and Mass Diffusion Equations 2.2.4.2 Crossflow Diffusion 4.1.4.2.4.1 4.1.1 Diffusion Convection 4.2.2 Governing Equations 2.2.3.2 4.5 SMAC Scheme and Compressibility CHAPTER 3: DIFFUSION MODELLING 3.2 .2 Energy Equation 2.2 Turbuletit Diffusion 3.1 Skew Differencing by Huh 4.3.3.2.3.3. Implicit and 4.1 Truncation Error Diffusion in a One Dimensional Problem Another Approach for a One Dimensional 4.4 ADI Schemes Conservation 4.2 Truncation Error Diffusion 4.3 Basic Assumptions and Limitations 2.3 Truncation Error Diffusion in a Two Dimensional Problem 4.3.4.
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
. numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig. 135 5. . by De Vahl Davis and Mallinson's formula along the line BD in Fig. .3(B) . 137 5. . (D+DDM). 136 5. 5. . .. .3(B) . Donor cell solutions with various diffusion corrections for a pure convection problem with 8 = 450 .15 . .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 .3(B) . .. 5. .. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with 6 = 63. .14 140 5.. . 144 . . . .. 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. . . . CA in Fig..430 . 5. .12 .. . . 142 5. .13 139 5. .43* .convection problem with 6 = 450 Comparison of the true solution and solutions by donor cell scheme.9 5. . problem with 6 = 63.11 Comparison of the true solution and solutions by donor cell scheme.8 .. .16 143 5. . True solution and donor cell solutions with and without diffusion correction for a pure convection . . skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure . 5. .
.. .. 5. . .. . .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 . Comparison of the true solution.. 145 5. . 153 . . . .. .19 . ...147 5.. ..18 Comparison of the analytic solution.21 S. . . 5. . 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.. .. 5. . 146 5. . .. 152 5. ..3(B) ... ..24 . .20 .19 .. 151 5.. . Simple recirculating flow field for test of the implementation strategies of the corrective scheme ..23 . . numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig... .. ... . . . . . .. . ..10 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. 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 . . 148 5.22 150 5... . . ..19 ..
154 5. . . .. Steam concentration distribution after one time step. .28 . .. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. with mesh point implementation of the corrective scheme .3 sec. with donor cell differencing of convection term without any diffusion correction ..33 Comparison of the ADI solutions for the profile of 4 at t = 10. . .. ..30 5. . . 155 5. . .0 sec . . . 0... .0 sec . .30 with . .31 1i60 5.27 .3 sec. different time step sizes Boundary conditions in terms of 4 for the problem in Fig.3(B) = e.. . 1 178 . . . 157 159 5. 5. . . .32 161 5. 5. . . . . .30 for At = 1.. . . .. 156 5.1I 5. 163 A.. . . Problem geometry for comparison of the explicit and ADI schemes .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 .29 Steam concentration distribution after one time step. 0. . . .30 for At = 0. with mesh interface implementation of the ... .. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. . 5. .5 sec and the Courant limit of 1. corrective scheme .. . . . .3 sec. . 5. . ..0 sec along the line AB in Fig. . Steam concentration distribution after one time step.cx T where 4 is given by. . 0.0 sec and the Courant limit of 1.
1981 . . Page . . . 2D. . . . . . .2 . . Comparison of the Von Neumann analysis and numerical experiments for the stability condition in terms of the cell Reynolds number in an explicit. . . donor cell differencing of the convection term . . . .114 . . . . . . .12 LIST OF TABLES Number 4. . . . . .1 List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research. 78 4. .
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 .C x y D D ah D as Dhs D .D y.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.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 ./Imaginary (1) Turbulent kinetic energy (2) Thermal conductivity (3) Ratio of specific heats.d xy e f Gr h I Im k g M Nu P Pe p y Az Pr Q Prandtl number Heat source .
tan.1 v u Mesh configuration. 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 .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.
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 .
( 1~111111_~. The error due to decoupling is negligible in a slow transient where the . 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. the response of the containment during an accident can be divided into two stages. The governing conservation equations are decoupled in order to simplify the solution procedure. Some remedies have been proposed to deal with However. Some simplifying assumptions are made concerning the thermodynamic state in the slow mixing stage.I~~U1i* 11*11044~1 1~. The distinct feature of the second stage is a much longer time scale in comparison with the first blowdown stage. The four components. From the hydrogen transport point of view. the fast blowdown stage and the slow mixing stage. air. hydrogen. it is necessary to understand such problems. the fluid dynamic phenomena in the containment in order to justify those remedies. steam and liquid droplets are assumed to be in thermodynamic equilibrium and the relative humidity is assumed to be 100%.__ 16 CHAPTER 1 INTRODUCTION One of the major concerns in the Three Mile Island (TMI) accident was hydrogen gas accumulation in the containment. although it may be less than 100% when there is no liquid component present.
c. k. The turbulent diffusion constant is predicted by the kc model [45). The turbulent kinetic energy. The diffusion constant. The molecular diffusion constant is predicted by Wilke's formula [73) for multicomponent diffusion and ChapmanEnskog formula [5) for binary diffusion. . Convection is assumed to occur as a homogeneous mixture. and the total diffusion constant is the sum of the diffusion constants of those two mechanisms. molecular and turbulent. 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. The turbulent kinematic viscosity. Finally. 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. resulting in the same convection velocity for the four components. of each component in multicomponent gas depends on the mole fraction of that component.17 state change over one time step is small. are determined by their own transport equations. the phase change is taken into consideration to maintain 100% relative humidity. Convection and diffusion are the central issues in physical modelling efforts of hydrogen transport. and turbulent dissipation rate. Diffusion occurs by two independent mechanisms.
Truncation error diffusion occurs in the flow direction while crossflow diffusion occurs primarily in the direction normal to the flow. The effective diffusion constants of the two errors are of the same order of magnitude. The leading issue in numerical modelling of convection and diffusion is to minimize the error that occurs in the numerical solution procedures. Truncation error diffusion is a one dimensional profile error and crossflow diffusion is a multidimensional operator error [68) of the finite difference equation. There has been much debate on the numerical diffusion and many schemes have been suggested for the past two .18 which is the diffusion constant for momentum transport can be calculated directly from k and E. 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 numerical diffusion has two different sources. Since the error usually appears as an additional diffusion. has been used to describe the numerical error in general. 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 term. truncation error diffusion and crossflow diffusion. The turbulent Prandtl and Schmidt numbers are assumed to be equal to one. however the latter turns out to be the dominant error source in most convection dominant problems.
mesh point and mesh interface implementations. 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. The schemes presently being used can be divided into two categories. skew differencing and corrective schemes. Raithby's [53) and S. The corrective scheme is inherently better than the skew differencing scheme in that it is conservative and does not affect the simple solution procedure. 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. have been tested for recirculating flow problems. 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. ADI and implicit schemes. However. The corrective scheme can be implemented with any of the explicit.19 decades. The mesh interface . 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. Two implementation strategies for the corrective scheme. although with different stability conditions. The conservative property is essential in a long transient problem like the hydrogen transport in the containment.
[49) includes all the Development Laboratory) [6]. . simulation results compared with experimental data. 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.
21 CHAPTER 2 PROBLEM STATEMENT AND SOLUTION SCHEME 2. 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. hydrogen generation gave a serious concern about the integrity of the containment in the TMI2 accident. In addition to the pressure increase due to the primary coolant.g. for dealing with this problem. The major safety concern is that of keeping the containment pressure below a certain level to prevent a large scale leakage of radioactive materials. it is essential to understand the fluid dynamic phenomena in the containment.. installation of ignition devices. There have been some mitigation procedures suggested. After a while the slow mixing stage follows the initial blowdown stage and continues for an extended period of time. use of flame suppressants and enhanced venting capability. the hydrogen has received much attention in the safety analysis of nuclear power plants. a large amount of steam and water will come into the containment in the fast blowdown stage increasing the containment pressure.. e.1 Problem Statement Once a lossofcoolant accident (LOCA) occurs in a Light Water Reactor (LWR). Thereafter. containment inerting. In order to justify the design of any mitigation system. .
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.g. . thermal nonequilibrium. laminar and turbulent flows. etc. Therefore. a good computational tool has been required to predict the hydrogen concentration distribution.. e._ __ _ __ _ I_ ~ 22 Since a numerical method is suitable for this purpose. some simplifying assumptions should be made to use the numerical procedure without impairing the acceptable solution accuracy.
6) Equation of state: p = f p (ii .2) Energy: p[e e + Vvh] + = VkVT + e fg (2. The physical implications of the conservation equations will be given with their basic assumptions and limitations.7) .3) Mas s diffusion: S+ V*vp i .23 2.t = VD. (2.4) Turbulence kinetic energy: Dk 1 t 3k 3 P 1 k ok x .T) (2.1) Momentum: pgI + d + V0 (2.5) Turbulence dissipation rate: D 1 _ 1t StE xk Dt P Txk 0E ] + C t*( 1 p k ui xk ouk au. Continuity: V'v= 0 pl + Vvv ] = Vp + Vv = 0 (2. I+ 3x 1 .2 Governing Equations The governing conservation equations are set up to describe the postLOCA fluid dynamics in the containment.Vp i + 4.xk k 2 v t BT C2 + kg k Pr 3x 2k k t (2.S + + t Bui P au k aui + ax~ ) x k X axk gk Pr t t 'T xk (2.
9) D.14) m = fk (PiT) (2. = D.10) v = vm + Vt (2. Im + Dit it (2.24 Constitutive equations: 'N = f (p .19) Note that 1.T) (2.18) p = Epi (2.12) (2.16) kt = . = f (Pi.T) (2. .t (2. The subscript i denotes the four components as .k + k t m Dim = fD(Pi'T) (2.11) k = .15) Dt = v t (2.13) D.8) f (v) (2.17) t = C k 2/C E (2.
i=14. v. (2.21) D4 = 0 (2. three momentum. one energy. . H2' steam.25 follows: i=l i=2 i=3 i=4 2. one continuity.22) There are eleven conservation equations. The turbulence equations are written in terms of the Cartesian components in tensor notation. T. there is one moreequation than is required.k C for a three dimensional case.i (4). 2. 3. four mass diffusion and two turbulence equations with ten primary unknowns. p.20) i = 2 = 0 43 + 04 = 0 (2. The mass diffusion equations are summed up for the four components. w. u. air.2. liquid droplet. Therefore.1 Continuity and Mass Diffusion Equations The continuity equation is redundant with the four mass diffusion equations and there is'an inconsistency between them. The phase change occurs between steam and liquid droplets and the diffusion constant of liquid droplets is equal to zero.
27 remain valid throughout the transient.24) Now it can be seen that the continuity equation is valid if and only if the following Eq.27) Consequently Eq.26 are satisfied. Since the concern is only for the slow mixing stage after a LOCA.(Ep i ) a .25 and Eq.) i + E 1 (2. ap at S+ Vp = E(VDiVp i ) 1 (2.26) It is a reasonable assumption in a turbulent flow regime that the four components have the same diffusion constant. Eq. .27. it is reasonable to assume that Eq. Therefore. at p= 0 (2. 2. 2.Vp.) + V(Ep Jt i = (V'D.25 and Eq. Vp = 0 (2.4 are consistent only if the temporal and spatial variations of the total density are very small.23) The sum of the phase change terms is equal to zero. 2. 2. 2. 2. 2.26 . 2.26 can be reduced to the following Eq.25) Vp i = 0 (2.1 and Eq.
is also of an approximate nature and the exact form is given in the following.iei) + V'[Ep. LeaPi LeQ.30) Eq.31) The thermal equilibrium assumption is incorporated in Eq. 2.29 denotes the energy change due to the concentration change in a given control volume.ih. [c pihi . (2.i + V.2 Energy Equation The energy conservation equation.28) The temporal term can be divided as follows. p + £e. 2.29 to give the following. E it p + e.3.2.31 so that the internal energy and enthalpy can be written in terms of the temperature. = VkVT (2.27 2. Be. . 2. 2. I at t + V [vp.29) The second term on the left hand side of Eq.ih i ] = V*kVT (2.2.i = V'kVT (2. It is clear that the major contribution to this term will come from the latent heat of the phase change. Eq.30 is substituted in Eq. ae. (p.
T h.28 (Picvi )~ +V"[vTZpic pi] + efg 9kVT = (2.1 = c Vl. 2.c . 2.3 and Eq. Pic vi T c p p.34) 2.33) h (2. = c p1 .T 1 Eq. . P (2 33) (2.T P P e Piei p ih p.32 become identical by defining average internal energy and enthalpy as follows. It is necessary to clarify these assumptions and their limitations for a safe use of the given physical models and solution scheme.3 Basic Assumptions and Limitations The governing equations are based on some simplifying assumptions about the physical phenomena in the containment after a LOCA.32) where ei i = eg4g + e 4) Sefg g e.
vt. 4. The four components are treated as being nearly incompressible and the phase change rate is moderately small. 3. 3. 5.1 Assumptions 1. The change of the diffusion constants. 2. .3. The relative humidity may be less than 100% when there is no liquid component left. Dt and k t over one time step should be small because the turbulence equations are decoupled from the velocity field calculations.2 Limitations 1. 2. with no slip. The turbulent Prandtl and Schmidt numbers are equal to one. The phase change rate should be moderately small because the solution scheme decouples the phase change term from the energy conservation equation. The four components all have the same convection The liquid droplets move with the gas components velocity.3. The temporal and spatial variations of the total density should be negligibly small because of the incompressibility assumption. 2.29 2. The relative humidity in the containment is 100%. The four components are in a state of thermodynamic equilibrium.
. 2. the obtained velocity field is substituted in the scalar transport equations. The phase change is assumed to be zero in this step.1Eq. The equation of state. continuity/momentum equations set and other scalar transport equations set. In the first step the SMAC scheme is used to get the convergent velocity field with zero divergence for every mesh. k. E. the conservation equations are decoupled into two sets.4 Solution Scheme Eq. p. Eq. . is used to update the reference pressure and reference density which is important in calculating the buoyancy force.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. 2. Pi(4). It solves the . the relative humidity may be less than 100%. The SMAC scheme is In the second step.30 2. When there is no liquid component left. explained in detail in section 2.7.7 are a coupled set of equations with primary unknowns. u.5. 2. 2. v. energy equation and two turbulence equations. In the third step phase change is taken into consideration to maintain 100% relative humidity.i Since it is too difficult to solve the whole system of the equations and obtain a consistent solution for all the unknowns. T. which are the four mass diffusion equations.
2. it is possible to modify the SMAC scheme to accommodate a slight compressibility effect.36 and Eq.e.37): . the SMAC scheme can be used only for incompressible flow calculations. The continuity and momentum equations are given in two dimensional Cartesian coordinates in the following. i.36) .ax 2. (Eq.37 as follows. Continuity: au ax av ay 0 (2.37)y (2. ay (Eq. Since the continuity equation used in the SMAC is an incompressible form. 2.35) xdirection momentum: au au + u+v at ax au ay 1 p ax + Pgx + a 2u 2 ax 2 2au + 2 2 (2. 2.36) ydirection momentum: av av ua av 1 + pg + a y x2 Nv 2 (2. which can arise with net inflow or outflow boundary conditions. However.37) Py The vorticity conservation equation is derived from Eq. the divergence of the velocity field vanishes everywhere..1_ _~ 31 continuity and momentum equations to get the velocity and pressure field in an iterative way.
37 as follows.32 a C 2 a2 + u+ V = v+ ) at . 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 (2.37 as follows.36) + ~(Eq.38) =ay ax The Poisson equation for the pressure field is also derived from Eq.41) where fn and gn are the explicit quantities at time step n. 2.36 and Eq. 2.+ y ax ay The basic idea of the SMAC scheme is to separate each calculational cycle into two parts.40) S (2.39) where av au D= . 2.x ay ax 2 ay2 where au av (2. ax 2 2. (Eq. In the tilde phase the approximate velocity field at time step (n+l) is obtained from Eq. .37): a2 = 2 ax 22 axay a 2 ay 2 (a2D + a2D aD +1 2 at Re ax 2 ay (2.36 and Eq. ay 2. the socalled tilde phase and pressure iteration. 2.
a guessed pressure field or the pressure field at time step n is used to start the iteration. V2p = f(u. 2.41 have the right vorticity. Therefore.39 that the pressure field is not right due to the fact that the divergence of the velocity field is not equal to zero. _ Pi+j2P +P iPij+ pij f(uv)  Dn + l D n D D Ax 2y t (2.O 0 (2. 2. the following relations hold after the tilde phase. 2.40 and Eq. 2. The equation of vorticity transport.38.39 may be rewritten as A finite difference form of Eq.44) Eq.44 is given in the following.43) ax ay It can be seen from Eq.45) . 2. The impor tant point is that un + l and vn+l calculated from Eq. shows that the vorticity is independent of the pressure field.v) D (2. au ay au av ax  (2.41 are still unknown.40 and Eq. 2. 2. Eq.42) av . follows.33 Since the pressure field in Eq. 2.
.34 The pressure correction formula is obtained by conn+l only and putting D. (6u) L =  pAx (6u) R = t6p pAx (2. no more correction will be made.36 and Eq. The velocity correction formula is derived from Eq. Once the zero divergence velocity field is obtained. S+ at V"(p) = 0 (2.48) The SMAC scheme can be extended to a slightly compressible fluid flow.47) ) The pressure correction is given in terms of the residual divergence for every mesh. 2.j) 26p( 1 Abx 6p = h + _) Ay 1 At( A + 1 Ay 1 Dt D (2. The slight compressibility means that the NavierStokes equations may be solved safely with the incompressibility assumption. while the continuity equation has the following compressible form.49) . equal to zero. sidering the cell (i. 2.46) (2.37 so that it does not change the vorticity implemented in the tilde phase.
: 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.52) (2.uniform over the whole containment. 1 . . 2.51) v p where m. 2. 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. Now the problem is how to converge to a predetermined nonzero divergence field. 2.49 the nonzero divergence may be derived as follows.p_ p at 1 min/V At n p n n (2.50) The two terms on the right hand side of Eq.51 and Eq. D = V'v 1 p p at V p (2. The associated physical assumption in Eq. This can be done with the following modification of the pressure correction formula of the SMAC scheme. V: total volume of the containment.35 From Eq.
48.6p = . The velocity correction formula remains the same as Eq.(DD 1 1 + ) 2At( Ax Ay (2.50.53) where D O is given by Eq. 2. . 2.
in the containment after a LOCA. steam and liquid components are transported by convection and diffusion in the containment. Although turbulent diffusion is greater than molecular diffusion by a few orders of magnitude. hydrogen. 3. the diffusion constant of each component is calculated separately by Wi~k's formula [73] in Eq. As collisions occur more frequently.1 Molecular Diffusion Molecular diffusion occurs by the collisions of gas molecules. The thermal conductivity and viscosity are also diffusion constants in a broader sense for momentum and energy transport. Convection occurs as a homogeneous mixture resulting in the same convection velocity for the four components. Diffusion occurs by two different mechanisms. the process Since there may be three gas of diffusion is also increased. air.37 CHAPTER 3 DIFFUSION MODELLING The hydrogen. components.1. air and steam. It gives the diffusion constant of each component in terms of . molecular and turbulent. 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. 3. which should be modelled independently.
stants are given as follows.: yi: Diffusion constant of the ith component Mole fraction of the ith component D.38 the binary mixture diffusion constants and mole fraction of that component. Eq. The binary mixture diffusion constants are calculated by the ChapmanEnskog formula. The property of air will be replaced The resulting diffusion con with that of nitrogen gas.AB where DAB: AB Binary mixture diffusion constant in [cm /sec] .2) poAB D.1) where D. 1Y A B DAB 1Y B yA DBA 1Y C YA CA C YC CB B YC DBC C D = C YB DCB (3.2.0018583 1 + M A MA MB (3.: Binary mixture diffusion constant between the ith and jth components. 3.. T3 D = 0.
4) np = np 2 where the correction factor 4 is given by.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. 3.3..5) 2 n .AB: Dimensionless function of the temperature and intermolecular potential field for one molecule of A and one of B. In order to calculate the para meters aAB and 0DAB in Eq.2. =n [ p n = [+ S n . The error range of the ChapmanEnskog formula is reported to be about 610% [5). For nonpolar gases OAB and ODAB are calcu lated by Eq. ap = np £ (o n + o p) EE e (3.3) CAB = AB For the pair of polar and nonpolar gases a correction factor is introduced to account for the polarity. the following relations are required. P n ]. OAB = (o A + oB (3. 3.lll~*. (3.
3.65 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.= 1.9035 = 4.5 are defined as follows.6810 x 10 .40 The parameters in Eq.6 T1. 5 = 4. The containment pressure is assumed to be1 atm.9492 x 10 D ah D 1 T .9 x 10 25 25 25 [cm3 3 [cm ] : N2 : an = 17. H2 a n = 7.6947 as .2 E/K = 380 0 K a = 2.6 x 10 The fitting functions for the binary mixture diffusion constants are obtained by the leastsquarefit method. t* .
__Li~n_~PIIII ~.lii.7) k. i (3.lllI_ 41 Dhs = 2.k. = n i=l y.8916 x 10 where T: D: [OK1 [cm 2 /sec] 5 T 1.10) Viscosity in [poise) Temperature in [K]) LennardJones parameter in [A) Dimensionless nTumber .8144 (3.67 x 10 where : T: a: Svi : o2 2 (3. 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). = 2.8) j=l where ij 8 Mj j ij M The viscosity of nonpolar gases may be obtained by the following [5).
c (3. 2 n Yi umix = C 2 i=l Yi + 1.11) for the Prandtl number of Pr = c (3. The effective diffusion constant is therefore the sum of the molecular and turbulent diffusion constants. 200 C There is another formula suggested for the viscosity of multicomponent gas in the following [5]. seems to be the kc model originally developed by Launder and Spalding [45).[5]. The ks model sets up the transport equations with some empirical constants for the turbulent kinetic energy. k.42 The viscosity of steam is.009 cp at 1 atm. and turbulent dissipation rate. The . There are several models suggested for Presently the best model calculating turbulence effects. E.385 nn y. 3.12) where c P is the specific heat per mole at constant pressure.2 Turbulent Diffusion The diffusion process is greatly enhanced by the turbulence of fluid flow.y RT k=l PMiDik k i There is a useful relationship nonpolar polyatomic gases. Wsteam = 0.
.15) .1 Derivation of Turbulence Equations [15. The effect of turbulence appears as the additional terms due to the product of fluctuations which do not necessarily cancel out.u. u. = u. + Pi Pi = Pi + p. 34.' etc. 29.= v ( Dt + ui j u P= Dt ax x 2 k6 ij 3x_) 3 i (3.u. + u.43 turbulent viscosity is given directly in terms of k and C. etc.13) (3.' For example. momentum and energy equations by decomposing the variables into mean and fluctuating parts and averaging the resulting equations. 3.' j T'= a t x.2. u. 45] The transport equations for the turbulent kinetic energy and turbulent dissipation rate are derived from the continuity. These terms are treated by the eddy viscosity concept of Boussinesq and eddy diffusivity concept which are given in the following. (3.14) .
3. A two dimensional cylindrical coordinate form of the kE model is introduced in Eq.' u is transformed to the turbulent kinetic energy equation by the contraction.16) . j=i.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 (3.2. 3.2 ke Model The transport equations for the turbulent kinetic energy and turbulent dissipation rate form the basis of the kE model.16. 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 ] . 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.44 The transport equation for u.
17) The suggested values of the constants are given in the following table. 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. 3. 3.0 o 1.17.44 C2 1.3 The molecular effects can be included in Eq. Since the kE model is applicable to fully . C 0.metry.09 C1 1. it is difficult to determine whether a region under consideration is in a turbulent or laminar flow.92 ok 1. = C pk 2/ (3.45 The turbulent viscosity pt is given in terms of k and E in Eq.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.16 as follows. 3.
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.5/(l+Rt/50)] C2 =2 where R 2 0 [1.3 exp(R2t ) (3.lit k + l ak aui i + t (ax k + xk auk axi axk e . Rt.0 .0 a ui ( 2 p ax ax 2 ) Dk Dt 1 p ak [(. The following kc model is a modified form by Jones and Launder [40. D Dt p ax k [(t Dt+ ) e k + C k i + k x ) au axk C 2 E2 2 t 2.0. it is required to extend the model to laminar flow regime or to switch to laminar models according to some criterion.19) = pk2/VE .18) In the above equations C 1 1 ok and a retain the values assigned in the ordinary kc model.16.46 turbulent flow.2vak 2() k 2 (3. Eq. while C1 and C2 are to vary with turbulence Reynolds number. CP = CP0 exp[2. 41). 3. The model presently being used is just to use the sum of the molecular and turbulent contributions for every region in the containment.
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)
3. is for a one dimensional case.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. Eq. . n+l i+f EXACT n+l i FD n i+f  4P n a n+g i + At [ u( x ) i+a (4. 1 < c < n 2 4 + i Ax 2 11 4 2 1 i+d < d < 1 The mean value theorem.x+a) and [y. af f(x+a. we obtain the error at time step (n+l).8) 3 @n + u()n + a( ax i+b ax 4 n+g _ n )n+g .7. 4. n .i+c T i+d ax where the following relations have already been used.a(?) ) 2(.y) + (f) af (4. 4.6 from Eq. n+g 4 i+ Ax n i Ax 1 xji+ n i+l .y+b) = f(x. 4.y+b].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.51 Subtracing Eq. It can be extended to a two dimensional form as follows.
8 to the following form.10 are at some appropriate points in the domain [xi . they may represent the errors over many time steps in a stable scheme. 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.A fAxAt axat 2 B utx(a+b) . C and E are truncation error diffusion terms while the terms B. E . . 4. 4. 2 n+l (4i ) EXACT  (4+) FD = i n+1 f )2 2fx 1f 2 ) (f 2 2 .tax .F (4.D +aAx(cd)Atax 2 +agt .C ax 22 +At g t .10 are based on the assumption of perfect information at time step n. D and F are time derivative terms which become negligible for slow transients.52 Now Eq. The terms A.9 reduces Eq. Although the error terms in Eq.xi+ ] and [tn tn+). 4. 4.10) The derivatives with respect to time and space in Eq.
The analytical solution of Eq. 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. 4.53 4.11) ax Consider a case in which the domain is divided into N equal meshes and the mesh size is Ax. cL (4.12) and C 1 =N cL 1 e C2 ecL 1 .11 is given as. 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.1.1 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. 4= Cecx + C2 where c = u/a . u ax a22 (4.
4.13 can also be solved with the given boundary conditions by reducing it to the following form. Equation 4. i u x i+l2. (4.13) where the velocity u is positive.15) Equation 4. .11 is finitely differenced using donor cell scheme for the convection term in the following. 4.14) The parameter P in Eq.16 is easily obtained from Eq. i+  i = (P+l)(4)  4)i (4.54 Equation 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. where P =u a  l i24 + ii (4.15 and solved for in subsequent procedures. Equation 4. 4.14 may be recast to Eq.i + 4i2 Ax = a (4. i+l i = (P+l)( 1 .40) .15.16) . P(4).
2 1 = (1+P) (4 0) 4' N 92 = N2 (1+P) 2 (  0) = N. 4. i  0 = 4.I_.17) 01 can be expressed in terms of 40 and 4N N N by Eq.16. 4. 4. 4.~LI LIII~.13.12 and numerical solution Eq.1 (1+P) N 1 (4. 4. we obtain the solution for 4.19 into Eq. 4. (1 0 ) is ) P( N ¢0 (1+P) given as follows. 4.41 = N ) (1+P) [(1+P) N1  2) (4.18) (l+P) + 1 Therefore.~l*IIIIX1~~^ ~ ^I ^~Xi 55 Then.LI II . (1+P) 1 .20) Now the analytical solution Eq.'N1 (1+P) N1  ) 0 N . as.11 and Eq.17.1 Inserting Eq. 1 (1+P) i . Both the analytical and numerical .20 have been obtained without any approximation for Eq.(1+P) P (1+P) 0 + N .19) (1 .
24) It is also possible to go through the similar procedures with a central differencing form of the convection term in Eq.25) (4.25.13. 4.56 solutions will be reinserted to the finite difference equation.13. The in analytical solution Eq.23) (4. i+l 2Ax il S S2ax 4) i+l 24i + i 2 i (4. 4.22 we can obtain the effective Peclet number Pe of the finite difference solution in terms of the real Peclet number P as follows. and the effective diffusion constant for the numerical solution will be obtained.25) . 4. 4. 4. e (il)Pe (ePe1)2 NPe .21 and Eq. 24 j. Pe = in (1+P) (l+P) iIp2 N (1+P) .20 is also substituted for 4.+1 i + ) il 0 (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. 4.12 is substituted for 4i Eq. Eq. in Eq. 4.13.1 (4. 4.22) Comparing Eq. 2 ii P ( (1+P) (1+) N _ 1 N 4).1 e Therefore.
26) where 2 < P < 2 The effective diffusion constant De can be readily obtained from the following relation. 4.26 may be a good indication of the truncation error diffusion occurring in a general one dimensional problem.24 and Eq. 4. 4.27) In general.57 The effective Peclet number for the finite difference solution of Eq. As in the previous section the velocity u is assumed to be positive and constant.25 can be derived as follows. 4. it helps understanding the origin of the truncation error diffusion. Eq. uLx D De Pe  (4.2 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. Therefore. Although this approach is approximate and overpredicts the truncation error diffusion. Pe = kn (2P) 2+P (4. 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 4.1. The convection terms in .
the diffusion is dominant over the convection.28 and Eq. 4.24 and Eq..e. i.4 and Eq. D uAX 2 (4.26. ae u( x where 0 <f < ae + u() ax i+  u~x a2 ) . 4. Pe also goes to zero resulting in no diffusion error in Eq. the convection is dominant over the diffusion. u i Ax i1 u i+ a4 u() ax u(a) ax (4.29 from Eq.28.5 may be approximated as follows. Therefore.29 hold only when the Peclet number is large. 4.31) ND .58 Eq. the truncation error diffusion constant for a convection dominant problem may be given approximately as follows.( i+f 2 (4. It can be shown that as P goes to zero.e.30) Both Eq. 4. 4.29) i+ The truncation error diffusion can be quantified by subtracting Eq. 4. i. If the Peclet number is small. the profile of 4 will be linear and the truncation error will be reduced to zero.28) Ax (4. 4. 4.
two dimensional. 2 u + v 2 + +( ) (4.59 4. Since the gradient of 4 in the flow direction is negligible in a convection dominant problem. u i 6x ilj + v i by ij_a(i+1 2 4x i + il1 +ij+l 2i 2 + 0iji by (4. 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. 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. one is the truncation error diffusion and the other is the crossflow diffusion. 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.32 is finitely differenced using donor cell scheme for the convection term. A steadystate.1. conservation equation with no source term is given in the following.3 Truncation Error Diffusion in a Two Dimensional Problem There are two sources of numerical diffusion in a two dimensional problem.33) .
x[cos + (cose tane 1 + sine) + sine tane ] ax xY (4.v( ay ) ij .n) ( = x cos6 + y sine I = x sine + y cose (4. then the truncation error of the convection term is given as follows.34) where tane = Ay/Ax tane1 = v/u The coordinate (x.35) .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 = .1.2 is applied to Eq.33. 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 . transformed to the coordinate ( .60 The approach in section 4.j+ + v(a) ly i+h. 4.y) is by rotation.
34 will be transformed to the coordinate system (t.36) where U = /u2v 2 Now Eq. 2 (A)  [(sine tane1 + cose) 2 + a2 (sine + cosetan6 1 ) a + (0) 4 an U" (4.32 is reduced to the following form.37) Equation 4.61 where 5 is the coordinate in the flow direction. dominant truncation error diffusion occurs When only in the flow direction.n) given by Eq. 4.37 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. 4. e is equal to 0 or 7/2. the problem is basically one dimensional and the truncation .35. 4. y v e Sx Then Eq. 2 2 2 + a = ( 2 ) (4.
1(A) is transformed into .l1(A). The hot fluid will come out of the top surface and the cold fluid out of the right surface. The origin of the crossflow diffusion is illustrated in Fig. which shows a single mesh with pure convection.68) and it was clarified that this is the dominant source of the error in most multidimensional problems.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. 4. In the donor cell scheme Fig. The crossflow diffusion comes from the multidimensionality of the problem. that will be named here as the crossflow diffusion.4. 4. It is entirely different Recently this addi from the truncation error diffusion. 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 4.62 error diffusion occurs in the flow direction as in a one dimensional problem. 4. tional false diffusion was explained by Patankar [52] and Stubley et al.1. [67. the hot and cold fluid enters the mesh from the left and bottom surfaces at 45* angle. In Fig. therefore it exists only in two or three dimensional problems when the flow direction is not aligned with the mesh configuration.3. 4.
4.63 Hot Mledium Hot 7 Hot Cold (A) Cold (B) Cold (C) Fig.1. Illustration of the crossflow diffusion in a single mesh with pure convection .
4. diffusion occurs in multidimensional donor cell differencing of the convection term. 4. . A 3 . in The numerical results Fig. 4. 2 (lp)[l+p+p + .&x u v ax Ay The effective diffusion constant for the crossflow diffusion will be calculated in the two dimensional Cartesian coordinate in Fig.1(B) are again interpreted as Fig. CAB which looks like a long onedimensional rod. and the sum of the values This distribution along the points A1 . In the donor cell scheme it is distributed all .64 Fig. .2 which shows a row of meshes aligned in the xdirection. . The value of 0 at the point A should reappear at the point X because there is no physical diffusion.3. an appreciable amount of false program user. 4. . of the value of 4 causes the crossflow diffusion. at those points is exactly equal to 4. 4. Another illustration of the crossflow diffusion is given in Fig. This is a pure convection problem Consider the fluid element The fluid with no physical diffusion. A 2 . ] = (lp)4 = where u .and intermediate temperature fluid will come out of the top and right surfaces. Then homogeneous mixing occurs in the mesh.1(C) by a Therefore.1(B) where the velocity components normal to the surface are considered.
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. Illustration of the crossflow diffusion in a row of meshes aligned in the xdirection with pure convection .2.
66 4. V A2 e A.3. 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. Geometry for explanation of the crossflow diffusion in a two dimensional Cartesian coordinate .
p4A and (lp)4A and appears at the points Al and A 2 . k2 = Ax sine/sin(el+8) k3 = Ax sine 1 /sin(e 1 +8) After substituting appropriate expression for each term. right and left hand sides.67 element was on the line CAB at time t in the velocity direction. which shows that the value 4A diffuses out along the fluid element. v. The travel of the fluid element is analyzed in Fig. The diffusion occurs both to the The diffusion to the right hand side is considered first in the following. sine cose cos= sx Uacos61 sin(O+e ) I DRHSS .4. the following right hand side diffusion constant results. Gradient) A /k 1 =A of 4 Current 2 2U of 4 Therefore. Diffusion A t1 . 4.t 0 PA £3 (Current) 2 1 constant ) where 1 = (Gradient) = U3 Lx/cose. and is translated The value 4A at the point A should reappear at the Point A'. In the donor cell scheme it is divided into two portions.
4.4. Geometry for explanation of the crossflow diffusion in a rotated two dimensional Cartesian coordinate .ct t2 t3 Fig.. Ya.68 ti .
39) Equation 4.69 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.38 is the effective diffusion constant when the diffusion is confined to the one dimensional fluid element.40) . sine cose RHS LHS cos1 sin(+8) (4. 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. There are two diffusion components in a two dimensional problem as follows. 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.39 is identical to the following simple expressions. The crossflow diffusion constants are the ratios of the current and gradient of 4 in the x and y directions. D D = uAx(lp) = VAy p (4.38) Equation 4.
De Vahl Davis and Mallinson [24] also derived the effective crossflow diffusion constant and their result is given in Eq. When both 6 and 61 are equal to .42 give the same result. Eq. D DDM = UAlxy sin2e= 3 3 4(by sin36 + Ax cos36) UAx tan6 1 sin26 3 4(tan8 sin6 3 + cos3) (4. 4. 4.42 underestimates the crossflow diffusion as shown in Fig. (4.6.41) Equation 4.40 and Eq. In Fig. 4. 3 0 when tan 6 = tan6 1 . < 6 < 81 or < .42) Equation 4. 4.42. that is to say.41 shows that DRH S (or DLH S ) is maximum at an Tr7LHS angle of 61 < 6 e between 4 and e1. RHS D = It can be shown that.38 should be differentiated with respect to 6 in order to find out the velocity direction with maximum crossflow diffusion. 4. Now the analysis will be extended to a three dimensional case in Fig. 4.6 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.70 where p = u Ax u v By Equation 4.5. In the donor cell p' y scheme the value 4 is divided into three portions px 0 .
5 80 \ % % 4 1. 0.0 150 300 450 600 750 900 Fig.5.0 0 0 .71 Cross flow Diffusion 2.0t 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 .5 I I 0. 4.
72 z C V Az X 0 o a X AY Ay B y x z C p . Geometry for explanation of the crossflow diffusion in a three dimensional Cartesian coordinate .6. p Fig. 4.
t4 cos8.43) Since the point X is on the plane ABC.43. £1 = AX =/(x4cosa)2 + (4cos)2 4cos) 2 = Ax 2 + 42 2Ax£ 4 cosa . £4 cosy) nB . along the directions XA. 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.44) 4 Therefore. n C denote The coordinate of the Any point on the plane ABC should satisfy the following Eq. point X is given as follows.45) The lengths of AX. B. S+ Ly Ax + Az = 1 plane ABC (4. XB. X(£ 4cosa. BX. XC. menon. S= 4 + y z zAx 1 cos+ + cos Ay Ax + cosy Az (4. 4. the following relation holds. cosa cosB cosy (4. and C instead of at the point X.and nA unit vectors along those directions.73 z occurring at the points A.
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 . nB.47) etc. 4 cosy] 1 nB = [i4cosa. Therefore. n C can be obtained in terms of their components in x.  n A = XA +1A  OA . nB' n C can simply be given as the products of the transported quantities and velocities. y. (Az 4 cosy)] (4. Z 4 cos8.OX (4. 2 o C 14cosy] = 1  £ 4 cosa. (Ay4 cosS). z directions.48) The currents in the directions nA .46) + The unit vectors nA .
75 I(Current) XB = Py p2 t I(Current) XCI where At = k /U z '3 At (4. (4.49) The gradients of 4 in the x./ Ay 4/6x Q/by (Grad)z = 4/Azl (4. Px D (Ax£ + Py D + nC(4 (Current)x 4 cosa) c osa) At ( 4 (cosa) (Grad)x = 4/Ax Therefore. y.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. z directions are given as follows.51) p Ax D x = lt (Axk 4 cosa) + p Ax . (Grad)x = 4./ Ax (Grad) = 4.(4 pz x co s a ) ct 4 + at ( 4 cosa) 4 .
p Ax D = ux[ £4cos x .1 = uAx(lpx) (4.76 p D x Ax At A2 At t cosa(p +p +p) 4 x y z SxA At Ax L cosa at 4 (4.52) Inserting the expression for At given by At = £4 /U. tensor viscosity method. finite element method. . Dy = vLy(lpy) Dz = wAz(lpz ) The rotation of (4. most important ones may be the skew differencing. we can simplify the result as follows. etc. This section is a review of the schemes that The have been used to eliminate the numerical diffusion.54) 4. indices will also give the same result.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.53) The crossflow diffusion constants in the y and z directions can be obtained in the same way. although none of them has been successful enough to be accepted widely.
the interpolation between neighboring points may still give some crossflow diffusion. Since the upstream point in the velocity direc tion does not necessarily fall on the mesh point where the quantity under consideration is defined.^ILli. but just appreciably lower than that of the donor cell scheme... Basically the inclusion of four additional corner points contributes tomore accurate numerical modelling of the convection. skew differencing and corrective schemes. In the skew differencing type schemes finite differencing of the convection term is done not in terms of the x and y directions.II1~LI1I~~ ICII1_1___~1___*_____11__ 77 Most of the schemes in Table 4. but directly along the velocity direction. The crossflow diffusion is not eliminated entirely in the skew differencing scheme. 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.~_ 1 i_... However. In the skew differencing scheme the 5point relationship becomes a 9point relationship. S. The corrective scheme is to reduce the diffusion constants in order to compensate for the additional false . the interpolation should be done between neighboring points to obtain the value at the upstream point. The donor cell treatment of the convection term results in a 5point relationship in a two dimensional problem..4 fall in the category of a corrective scheme.3.1 can be categorized into two types.
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 .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 .j. List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research.1QUICK t ETovh ci .Highe cdtr upwi:nd :"'ie Cdi't. upwind f~iite di Teretn atl.aptjve me.inte Conribo:r/me. 1981 [65) .1.thod LLxW'en doff  21 x11 80 x 400 Table 4. viscosity method Sel.78 Chss' iction Cen:ered . Grandotto Hickmott Firs. Esposito Glas and Rodi Huffenus Lnd Khali.thod of chaaceriics Fiite anuryic method Tenso.e elements Clifft Dontcl t al. Wilkes Elbahu ct l.ronLUD Lillinion/VUDCC Nuana./LLUE Huint/QUICK Wilkes/LUE Lillimpon/SUD LiiL.ordc.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 !. £Ebzhau t aL/PATANKAR Lilinpion Orlandi Priddin Wada cit l..ad.hod Mesh *txLnsformation rr.
54.79 diffusion. D. y and z directions will be reduced by the amounts of D .53 and Eq. 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 tensor viscosity method is one example of the corrective scheme. and D z in Eq. where the balance equation is set up for a given control volume by considering convection and diffusion at the control surface. 4 = C1 + C2 n = C1 + C2 (  (4. 4. The correction formula The diffusion con used here is Eq.3. It is therefore a conservative scheme.1 Raithby's Scheme Raithby's scheme [53] is an Eulerian type of approach. although the validity of its correction formula is not clear.55) where n is the distance normal to the flow and U = /u 2 +v 2 .. 4. the corrective scheme can give a numerical solution which is almost free from numerical diffusion. The corrective scheme is valid only if the effective numerical diffusion constant can be predicted accurately. 4.53 and Eq.54. 4. 4. Since the crossflow diffusion is dominant over the truncation error diffusion and the effective diffusion constants for the crossflow diffusion can be predicted theoretically. stants in x.
7. control volume and notations in Raithby's scheme . the calculation domain showing grid lines. A segment of. 4.80 Fig.
Although this scheme is nonconservative.56) where U = /u is the coordinate in the velocity direction and 2 +v 2 . The convection term is differ enced in the velocity direction between mesh points.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. which may be too complicated and time consuming for a practical purpose. 'ij US (4.57) where the upstream value €US may be calculated by .3. not in terms of the convection that occurs at mesh interfaces. This should be repeated for the four control surfaces of every mesh in a two dimensional problem. 4. The inclusion of the four additional points also complicates the structure of the coefficient matrix in an implicit scheme. it is much simpler than Raithby's scheme and becomes conservative in a unidirectional flow.2 Skew Differencing By Huh Skew differencing scheme here treats the convection term in a Lagrangian way.= U ax dy SC (4. u 4 + 'v.
58 for the numerical diffusion constant is apparently not consistent with the results in Eq. 4. 4. DDx.53 and Eq. D are derived .3. 4.54.58) t2U= Eq. z directions to compenx sate for the crossflow diffusion effect. D . 4. uv (4.2.82 interpolating two neighboring points.3.3. DD y . the crossflow diffusion constants D in section 4. The validity of this scheme is based on the fact that the crossflow diffusion can be predicted accurately. DD z in x.3 Tensor Viscosity Method The tensor viscosity method treats the numerical dif fusion constant as a tensor quantity. y. Dx = uAx(lpx) D Yy = Ay (lpy ) The expressions for .2.4 Corrective Scheme By Huh The corrective scheme is to use the reduced diffusion constants. 4. Some of the results of this skew differencing scheme is given in section 5. the tensor viscosity [25) T = Lt uu u2 The expression for is given in the following.
After the crossflow diffusion constants are assigned to every mesh point. The velocities at two boundary faces are averaged to get the representative velocity.+ . they should be averaged again between neighboring mesh points to calculate the crossflow diffusion current at mesh interfaces. .1 Mesh Point Implementation The mesh point implementation is to assign the crossflow diffusion constants Dx .4.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.. 83 D z = wAz(lp z ) where px = u (4. Dy . which is used with the cell dimensions to get the crossflow diffusion constants. the mesh point implementation and mesh interface implementation. In a coarse mesh or recirculating flow field the latter implementation strategy is definitely preferred. 4.Dz at the center of a mesh where all quantities are defined except the flow field.59) v u . Two implementation strategies have been tested so far._IIU__lpULI~~I___~lLlt i~Lli ~i ~iL~pn~sl.3.+ w Ay Lz x etc.
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.8. The two velocity sets (Ul. as shown in Fig. For the first two cases the cross flow diffusion constants are zero because the velocity vector is aligned with the mesh configuration. can be directly used with cell dimensions in Eq. The third case is expected to have a flow split as shown in the figure. For the second case the two outward velocities. There are three possible cases that should be treated independently in a two dimensional case. The right interface will have the crossflow diffusion constant (Dx1+Dx2 ) and the top and bottom interfaces. and the resulting D x and D y will be assigned to each interface.V2) are used to give (DxlDyl) and(Dx 2 'Dy2 ). so that they are proportional to V1 and V2. 4. The velocity U is split into two components.84 4. 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. The mesh interface implementation is always recommended although it has a more complex program logic than the mesh point implementation. U and V.Vl) and (U2.4. Dyl and Dy 2 . 4. . The same logic can easily Appendix C includes be extended to a three dimensional case. U1 and U2.3.59.
_~I~~ _~~~~_I__~__~~ __ 85 (1) (2) (3) Ul U2 V2 VI U U V14V2 VI+V2 V2 U2 = UV2 U = Ul + U2 TV2 Fig. 4. Three possible flow configurations for mesh interface implementation of the corrective scheme in a two dimensional problem .8.
conservation physical constraint accuracy stability 2 ax 2 2 y 2 +ux +at + 7 + T7) + S sz (4. 4. Explicit. 4.61) 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. Eq. 2.60.62) . The pros and cons of the three schemes will be compared in the following four viewpoints. Implicit and ADP schemes There are three typical solution schemes. for the general conservation equation.) = i1 n D Ax ( i+1 2i l n n + n ) + iS i1 (4.86 the fortran program for calculating crossflow diffusion constants in an arbitrary flow field. 1. which is a parabolic partial differential equation.60) Explicit (lD): n+l I n i At + u (i i x n i. Implicit and ADI (Alternate Direction Implicit). 3.4 Comparison of Explicit. 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.. 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. 2 Ax n+l 2 * ij v n+l ) ij n+l Ay 13j1 n+1 ij+l n+l 1j1 At Ay2 + S (4.63) 4.4. 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). + 4).
there are some nonconservative solution schemes like ICE (Implicit Continuous Eulerian). the conservative property may not be crucial. A scheme may be both stable and nonconservative and also both unstable and conservative because the conservative property is independent of the stability. where the convection terms of momentum equations are linearized in a nonconservative way. In order to guarantee conservation. + CI 4. Although they usually give . For a long time transient. implicit and ADI schemes can be made conservative by using appropriate differencing forms for the convection term. When we are concerned with a steady state solution.88 the domain boundary is equal to zero. however. All the explicit.2 Physical Constraint The physical constraint is closely related with the stability in a practical sense. 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.4. Although a conservative form is preferred to a nonconservative one.
2. 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.4.89 similar constraints on the time step size and mesh spacing. they are from entirely different origins. )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.1 Diffusion There is a maximum net flow that can occur between neighboring meshes by diffusion. new values of 4 are given as follows. 4. The laws of physics impose some restrictions on the numerical scheme to get physically reasonable solutions.D . R C AxAy = CAxy + D R Ax Ax CAy C y At At ¢4AxAy = 1RAxAy .
R AxAy(4) C) > 2D (R Ax CyAt 2 DAt X2 < 1  (4. scheme given in Eq. we have the following physical constraint. Both of the oscillation and instaThe ADI bility should be avoided in a transient problem.64 may be extended to a ddimensional case as follows. 4. d 2D t < 1 (4. should still hold. .65) Ax Equation 4. 4. however. 4. there will be a damping oscillation in the solution when the time step size is greater than that given by Eq. When the physical constraint is ignored and the time step size greater than that given by Eq. t' > Q' . the solution will be unstable or stable with damping oscillations.65 is used.65 may be a too conservative criterion in a real problem.64) Eq. Therefore the physical constraint should be respected in addition to the stability condition to get a meaningful transient solution.64 or 4.65.90 Since the relationship.63 is unconditionally stable. 4.64 or 4.
66) When a time step size greater than the Courant limit is used.2. a portion of the control volume C will cross the interface S. It is based on the assumption that 4A is the repreIf At is greater sentative value of the control volume A. than the Courant limit.4. Therefore. Then uAttay is no more an appropriate expression for the convection quantity during At. 4.4. At < 'x u (4.B 4y S The explicit and donor cell treatment of the convection term gives the convection quantity at interface S as ulAAtAy.2 Convection C A u B C A ). the damping oscillation or instability may occur as in the case of diffusion.3 Accuracy Realistic interpretation of a finite difference equa tion comes from an integral form of the conservation .91 4.
4. I I t i r _ I I [ n n+1 n n+l _ 0 It Explicit Implicit CrankNicholson $i+. ~~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.9. Profile assumptions of Q in space and time for various differencing schemes .92 n+l n+l 4' I I.
Since the exact profile depends on the Peclet and Strouhal numbers of the governing equation. The Von Neumann .4. whether the error introduced at a certain step will increase or decrease as the calculation goes over to the following steps. accuracy depends on how well we can compute the convection and diffusion quantities at control surfaces over the time At. the error propagation is suppressed and the largest error contribution will be from the previous time step. 0(Ax 2 ) is The also not appropriate for finitely large At and Ax. The infinitesimal approximation of differential terms is not appropriate in most problems of our consideration. The accuracy expressed in terms of O(At). 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.4 Stability The stability is a mathematical problem. The stability can be checked by the theorems about matrix eigenvalues or Von Neumann analysis. the accuracy depends on the profiles 4 in time and space because the exact profiles assumed for are not known a priori.93 equation. the accuracy is usually the least important among the given four considerations. It depends on the maximum eigenvalue of the iteration matrix. 4. If its absolute magnitude is less than one. In other words. However.
but its algebra may get complicated.1 Explicit Scheme The Von Neumann stability analysis of a finite difference equation is simple in its idea.68) . 4..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. 1 4n = E~nei where I = .4.94 analysis is to expand the solution of a finite difference equation in Fourier series and check the decay of each mode separately.4. implicit and ADI schemes using the Von Neumann analysis. is expanded in Fourier series as follows. lowing sections. 4.67) The solution 4.4.1. 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 . (4.2 2i + D i+1 i n 2 i i1 (4. stability criteria will be derived for the explicit.4.
68 is substituted results.69) Dbt Dt 2 Tx.72) In order to get the stability of Eq.95 The absolute value of should be less than one for every When Eq. 4.70 may be rewritten as follows by defining cos6 x as t. Therefore.2dx(cosex) 2 + [Cxsinex 2 < 1 (4. [i . the function . the following expression for = lCxsin xI where cx d x uAt A tx  2d (lcos x) (4. f(t) = i 12 = (4d2_ 2 )t2 + (Sd 2+4d )t + C 2+4d 24d (4. 4.70) Equation 4. 2 11 = [1 .67.67. mode to achieve stability. 4.2dx(t) where t = cose 1 < t < 1 ] 2 +C 2(1t 2) 1 (4. in Eq.71) The function f(t) will be defined as follows.
73 can be solved by a graphical method using the characteristics of a quadratic equation. Therefore. f(t) < 0 (4.10(1). D(Discriminant) = (C 2 .75) The parabola f(t) may have three distinct shapes according to the sign of the coefficient of the second order term.96 f(t) should be less than or equal to zero for any t in [1. f(1) = 0 (4. . The parabola f(t) always meets the taxis because the discriminant is always greater than or equal to zero.1] Equation 4.1]. 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. (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.74) One of the two meeting points with the taxis is the point t=l because f(l) is equal to zero.2d )2 > 0 (4.73) for any t in [1. 4.
f(t). according to the sign of the coefficient of second order term . 4.. Three distinct shapes of the parabola.97 f(t) f (t) f (t) 'I I 4. r I I t i t 1 1 t 0.3 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..10.
2d Axis of symmetry: t = 4dx dx > C 2 x > .98 f(l) = 8dx(2dx1) < 0 0 < d < (4. 4d 2 C 2 4d C > 0 < 0 : : 0 < d dx > < 2 Cx The stable region in the coordinate system (Cx.1.78 is a finite difference equation with donor cell differencing of the convection term. 4.76) In case (2) f(t) is a straight line and f(l) again should be less than or equal to zero. . t=l as shown in Fig.76 and Eq. 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. 4.10(3).4.77) Summing up the results in Eq. 4.77.11.C (4.4. 4.2 One Dimensional Donor Cell Differencing of Convection Term The following Eq. we get the following stability conditions. 4. 4d 2 . 4.dx) is shown in Fig.
4.99 d x 1 1 0 1 dx = C2 x 2x C Fig.d ) for an explicit scheme with central differencing of convection term in a one dimensional problem .11. 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)
104 Im \ . .13. 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 .1 SRe 2 al +2 1+a +a 2 Fig. 4.
14. . .105 . Fig.. should exist for the stability of an explicit scheme with donor convection term in cell differencing of a general two dimensional problem . The region in the complex plane where the amplificatior factor. 4.Re Both al and a 2 should be in the shaded circle to ensure that Ial+a 2 +1I< i.
the stability However. At. where the points (Cx. the stable time step size for a given mesh spacing.90) a 2 = C [1 y R R Y + R (4. 4. 4. which depend on both the mesh spacing.89 can be solved by the graphical method as in the previous sections.e Y) (4. When Ax and At are already given. Therefore. The stability condition is again given by Eq.88 and 4. At.e Rx .4.d) should be in the shaded region for stability. Ax.15. 4.88 and Eq. 4. the stable range of Cx should be determined in terms of the cell Reynolds number Cx/dx'. I 1 2 + (1+)e Rx Rx 2 + (1+)e a1 = Cx [1 + Ie 1 3 .dx) and (C y. 4.106 Both Eq. Cx and d . which is independent of the time step size.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. The stable region shrinks on a linear scale with dimensionality.91) . A three dimensional case can also be solved through the same procedures and the results are given in Fig.1.16. and time step. The results are given in Fig.89 and al and a2 may be rewritten as follows.4. our usual concern is can be checked directly.
.15.dy) fcr an explicit scheme with donor cell differencing of convection term in a general two dimensional problem. Stability cc.dx) and (Cy.idition in the planes (Cx. 4.107 d (d ) d = x 2 1 x + 1 4 C x(C ) 16 d= C2_ C x 2x x Fig.
16.dx) etc.(C Cx x y or Cz) z d ) Fig. for an explicit scheme with donor cell differconvection term in a general three encing of dimensional problem . 4.108 d xy(d or d z ) Cx d dx + 1 1 Cx (C d! S3 2 . Stability condition in the planes (Cx.
) RX (1cose )) x + [C sine] x x < (4. 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. 4.92) Equation 4. (1) Cx 2 K2 1) > 0 concave upward .1] f(l) = 0 (4.90.95.__LI___IIII__LI*X__ ID I 109 The following condition for Cx and Rx can be derived from Eq.92 can be simplified as follows.95 ) The function f(t) satisfies Eq. 4. f(t) = C (1+K) 2 (lt) .94 ) D(Discriminant) = C 2 (K2C ) 2 > 0 (4. and has distinct shapes for the Following three cases.88 and Eq.93) ) where 2 K = 1 +.94 and Eq. 4. [ .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.C (+2x .
17.97 are shown in Fig.97) The results in Eq. 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. 4.96 and Eq. 4. 4. therefore the subscript x may be replaced with the subscript y in Fig.17.yu < 1 (4. The stability condition in Fig. V f . 4. R > 0 x : 0 < C < x 2 2 2(1+) x and (4.17 is applicable to all general 2D problems. (4.98) Ax .110 (2) C 2(K 21) = 0 linear (3) 22 C 2(K 2_1) < 0 concave downward For case (1) and (2). t=l. 4. to zero. The same result will be obtained for the ydirection. although it can be relaxed if there is a constraint on the velocity direction and mesh spacings such that the following' relation is satisfied. solution procedure is similar to those in the previous sections and the results are given in the following as.96) R 1 x < 0 < C x < (1+2) Rx .
111 x (C y C x  2 1 2 (1+ L) R x C = 2(1+ ) x 2 0 R (R x y Fig.17. 4.Cx) and (Ry.Cy) for an explicit scheme with donor cell ditferencing of convection term in a general two dimensional problem . Stability condition in the plane (Rx.
4. Stability condition in the plane (Rx.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.Cx) for an explicit scheme with donor cell differencing of convection term in a two dimensional problem with f v u/Lx .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. 4. 4.2 shows the com 1 l+f in Fig. (4.102) dx > . 4. 4.uLx. the stability condition is found to be the following. parison between the analytical result in Fig.2 Implicit Scheme The following Eq.Cx d > .4.100) For a general two dimensional problem. 4. and it can be seen that the implicit scheme is unconditionally stable if the diffusion constant is than . 4.4.17 should be Table 4.113 Then the number replaced with on the C axis in Fig. d >  greater Cx (4. 4.C Y y .99) (4.19 and Fig.101) (4.17 and numerical experiments.18.99 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.99) The stability condition is shown in Fig.20.
180 0.120 0.333 83.499 0.540 0.660 0. R x 0. 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 .380 0.2.3 8.260 0.114 Maximum Stable C.660 0.333 833.167 2.720 0.147 0.403 0.629 2.333 4.833 8.083 Von Neumann Analysi s 0.060 Table 4.780 0.120 0.488 0.019 Numerical Experiments 0.
Stability cndition in the plane (Cx.19.115 d =  x d =. 4.dx) for an implicit scheme with donor cell differencing of convection term in a one dimensional problem .C x 2x x 2x C x dx 1(C2+C 2xx ) Fig.
116 Cx=1  2 Rx 2 1 0 R x R C x R +2 Fig. Stability condition in the plane (Rx. 4.Cx) for an implicit scheme with donor cell differencing convection term in a one dimensional of problem .20.
103) jsj <1 In order for I.jto be less than or equal to one.4. The Von Neumann analysis of the first formulation in Eq. the following condition should be satisfied. 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.+d )] C ) {[1+(lcosex +dx)) +[fsinex }{[l+(1cosy) (+dy) + Y[siney } (4.63.117 4.4. One of the formulations has both the x and y direction components in each step. 2 _C C 2 {[(cos )(+d X))] C [sine I2} {(1cos6 C 2 )(.104) . 4. 4.3 ADI Scheme There are two possible formulations of the Alternate Direction Implicit (ADI) scheme given in Eq. C d x>2 (4.63 gives the following stability condition.
118 The same condition should hold for the ydirection.63 has the same stability condition as the fully implicit scheme. C d x > 2 2x (4. 4. 4.106) C d y  >  ~ 2 (4.107) Therefore the second formulation of the ADI scheme in Eq. C d y>  2  4 (4.105) The stability condition of the second formulation in Eq.63 can also be given by the Von Neumann analysis as follows. .
density is determined by temperaSince the natural convection of air ture and pressure. For example.1 Natural Convection Results The containment is modelled as a two dimensional rectangular compartment with an obstacle as a heat sink. Natural convection occurs due to the heat transfer between the obstacle and containment air. 5.1. the ADI scheme is explored to use a time step size larger than the Courant limit to save the overall computational time. 5. Finally. 5.1 and Fig.2. energy convection and heat transfer models are considered and the resulting flow fields are given in 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. Updating of the reference state.1 Updating of the Reference State The state of air is determined by any two independent properties. 5. 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.
4 sec AT = 100F T .8 sec .0 Btu/lbm 0F (Q ob = 208.29 Ibm A= 2 ft2 T. 0. = 680F air T = 580F Fig.5 ft/sec h = 1.8537x10tu/ft20 F sec (Cp) = 1. 5. = 677. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 10.1.120 p ).
8537x10 tu/ft2 F sec p)ob = 1. T ab = 580 F Fig. 5.4 sec 1 0 AT = 10 F T air = 680F .121 > 0.5 sec . Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 62.29 ibm (K 2 A= 2 ft T. 5 ft/sec h = 1. = 677.2.0 Btu/lbm OF (c )b = 208.
122 gradients. Told' Tnew: Temperature of the fluid at old and new time steps.total Pold Told new old Qcell cv Mf. however. temperature and pressure are equally important parameters in determining the flow field. Pold' Pnew: .1) T Pnew = old + Ttotal c v M f. time Pressure of the fluid at old and new steps. The density change of liquid. depends primarily on the temperature. not on the pressure because of a high modulus of elasticity. new Pnew RT new (5. The equation of state of an ideal gas is given in the following.cell where Pold' Pnew: Density of the fluid at old and new time steps. c : Specific heat of the fluid at constant volume.
5.0 Monatomic. the atomic gases have different values of k. k = c /C for liquid and 1. 5. 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. .4 when internal energy is used for enthalpy in air flow. 5.. The value of k is 1. An experimental correlation is given in Eq.1.cell. diatomic and polyTherefore.4 for air. . The Nussett and Grashof numbers here are based on the axial length of one computational mesh.3 Heat Transfer Modelling The heat transfer to the obstacle is modelled as a natural convection from a vertical surface.1. convection energy transfer should be in terms of enthalpy instead of internal energy.123 Mf. Mf. but they may also be based on the vertical height of the obstacle.total : Mass of the fluid in one cell and in the total domain. The energy convection will be underestimated by a factor of 1.2 in terms of the dimensionless numbers.2 Energy Convection Another difference between liquid and gas is the ratio of the Specific heats. Qcell total : Heat input to the fluid in one cell and in the total domain.
8537 x 10 .3) Pr = 0. .124 Nu = C(GrPr) n ( 5. L = 2 ft AT = 500 F p = 0.876 The heat transfer coefficient is given as. h = 1.sec .021 n 0.2179 x 10 B = 1.2) 3 2 Gr = gS(AT)L p 2 (5.4 The following data are used to calculate the heat transfer coefficient at the obstacle wall.708 (air) GrPr 105 .555 0. Gr = 9.2 ft/sec' 5 V = 1.4 lbm/ftsec Btu/ft 2 F.923 x 103 Therefore.25 0.0763 ibm/ft 3 g = 32.7134 x 108 Nu = 89.109 > 109 C 0. This may be a typical value for the natural convection heat transfer in the containment without any phase change.
Heat transfer area between the obstacle and fluid over one computational mesh.1. pf Pob: c h : Fluid density.4 sec. Heat transfer coefficient between the fluid and obstacle.4 Turbulence Modelling Laminar flow is assumed in obtaining the results in Fig.1 and Fig. If the flow is turbulent.125 The VARR input [21) requires the time constant of the heat transfer instead of the heat transfer coefficient. Obstacle density.2. 5. The time constant. T. the heat . Specific heat at constant pressure. Therefore.4) where CQ is the source term in the energy equation of fluid. 5. (Mc )ob (MC) obf hApo b ob f (5. can be obtained from heat balance between the obstacle and fluid as follows. T = 677. CQ = Cpob (TT) (5.5) where M A : Mass of one computational mesh. 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.5 show that the numerical solutions is close to the analytical solution and the truncation error can be neglected.2 Numerical Diffusion The truncation and crossflow diffusion errors are compared in a simple geometry where an analytical solution can be obtained. 5. 5.6 and Fig.3(A) is parallel with the mesh orientation so that no crossflow diffusion occurs. 5. The flow in Fig.2.1 Truncation Error and Crossflow Diffusion The truncation error diffusion is compared with the crossflow diffusion in a two dimensional. steadystate problem in Fig. is compared with De Vahl Davis and Mallinson's in predicting the magnitude of crossflow diffusion. 5.7 show that excessive numerical diffusion has occurred due to the .to the obstacle will be enhanced and the diffusion transport of momentum and energy will also be enhanced in the containment air. 5. 5.4 and Fig. The corrective scheme is tested in recirculating flow problems for mesh point and mesh interface implementations. The results in Fig.3. The flow in Fig.126 transfer. The skew differencing and corrective Huh's formula schemes are tested for various problems. 5. Figure 5.
for evaluation of numerical diffusion .0 B T specified Fig.0 (A) (B) AB=BC=5. 0 ST By =0.127 (A) .0 ay AB=BC=5. (A) parallel flow and (B) cross flow.0 u = 10.0 BT S=0.3. Problem geometry for the two cases. 5.0 x BT  @x a( aT + 2 2 a2 ay ) Th= 10.0 a= 1.07 (B) =0.0.0 u = v = 7.0 B T_ 0.
5. 5. spatial coordinate Fig.128 Analytic solution Numerical solution with Donor Cell T 10.4.3(A) . Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 10x10 meshes along the line CD in Fig.
~ I~.~ 129 Analytic solution Numerical solution with Donor Cell C spatial coordinate Fig.5.__~ . Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CD in Fig.~l r.__LL l. 5.3(A) .^XI ill I~LX_L~LI. 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. Comparison of the analytic solution.130 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Df cf T 10 spatial A coordinate Fig. 5.3(B) . 5.
.7.131 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Dcf T 10 / I / / . 1/ 5.3(B) .3(B) s' / line CA in ../ 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. Comparison cT the analytic solution.J 0 I' C spatial A coordinate Fig. 5. 5.
6 are for 5.12 for a simple pure . Figure 5.10 by De Vahl Davis and Mallinson's for the problem in Fig.9 is the result by Huh's formula and Fig. 5. Raithby's and Huh's.6. 5. 5. 4. Since the corrective scheme is based on the validity of those correction formulas.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.8.7 than in Fig. Figure 5.132 crossflow diffusion. Eq.3 Skew Differencing Scheme Two skew differencing schemes.54.2. 5. 5.810. meshes. 5. 5. 5. 5. it is necessary to test them for various cases.5 and Fig. 6 = 6 = 450. The problem geometry in Fig. more diffusion has occurred in Fig.8 has the arbitrary angles of 6 and 61 such that 6 = 600 and 61 76.2.6 and Fig. 10 x 10 meshes and Fig. 5. 4. 5. It can be seen that the former result is much better than the latter. Figure 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.4 and Fig. De Vahl Davis and Mallinson's and Huh's formulas give the same crossflow diffusion constants for the case.53 and Eq. 5.11 and Fig. are tested in Fig. solution.2 Validation of Huh's Correction Formula The crossflow diffusion can be predicted by Huh's formula.
7674 DH u h = 3.810 8 = 600 8 Meshes 24 Meshes Fig.2406 by = 1.0 Ax = 0.6603 DDM= 0.0267 u = 5.133 Yo 2 T=0 D a = 1.0 v = 8.8. 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 .335 Huh A e1 = 76. 5.
5 0 spatial coordinate Fig. (D+Dcf). 5.134 Analytic solution with D+Dcf Numerical solution with Donor Cell T 10. Comparision of the numerical solution with donor convection term and cell differencing of analytic solution with increased diffusion constant.8 . by Huh's formula along the line BD in Fig. 5.9.
5. by De Vahl Davis and Mallinson's formula along the line BD in Fig.10.8 .135 Analytic solution with D+DDM Numerical solution with Donor Cell T 101 spatial coordinate Fig. solution (D+DDM). 5. Comparision of the numerical solution donor cell convection term and analytic differencing of with increased diffusion constant.
25 65.5 68. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with e = 45* .5 0 6.75 81.136 Donor cell 100 100 100 100 100 93. 5.75 25 U 12.25 18.63 87. Comparison of the true solution and solutions by donor cell scheme.38 50 31. 0 100 100 100 100 100 50 100 50 0 50 0 0 V Fig.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 34.25 0 Skew differencing by Huh (True solution) 100 100 100 100 50 0 0 0.
9 33.52 1.25 1.430 0 50 0 Comparison of the true solution and solutions by donor cell scheme.96 17.01 55.137 Donor cell 100 100 100 100 50 .85 0.99 10.33 70.12.E 21 0 12 True solution 3.25 0 0 0 .430 .67 5.70 4.8922.25 53.83 72.13 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 6.59.6.5C 100 100 50 Fig.22 9. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure covection problem with e = 63.56 1.26 80.2! 90.91 71.A Skew differencing by Raithby 100 100 100 100 96.50 2.91 31.6 0 Skew differencing by Huh 100 90.74 38.47 23. 62.11 3.86 16.11 0 11.43 7.3E 21.71 20.5 0 0 63.3: 11.88 81. 5.3' 40.1 6.25 37.5E 25.
3. Raithby's scheme is an Eulerian Therefore.2. Figure 5.13 and Fig. Some crossflow diffusion has occurred in Huh's scheme. 5. method while Huh's scheme is Lagrangian. 5.138 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. but less than inthe donor cell scheme. 8 = 1e= 45*. to obtain the true solution by subtracting the additional crossflow diffusion .11 has a flow given in sections 4. The purpose of the corrective scheme is.3. The details of those schemes are Figure 5. however. both of them are conservative in a unidirectional flow. former is conservative while the latter is not.14 show that the analytical solution can be reproduced by Huh's skew differencing scheme for the case.12 has a flow at a 63. the However. It is indicated that Raithby's scheme may give unphysical results in problems with steep gradients of the quantity under consideration.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.430 angle and neither Raithby's nor Huh's scheme can reproduce the true solution. 5. Fig.1 and 4.2.
3(B) .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. 5. 5.139 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh C A spatial coordinate Fig.
5. 5.3(B) .14.140 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh 10 5 0 spatial A coordinate Fig. 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.
18 show that the crossflow diffusion can be eliminated by the corrective scheme and that a finer mesh spacing always gives a better solution.19 is a hypothetical 3 x 3 recirculating flow field where the inlet boundary values are specified on the left and bottom surfaces. 5. The corrective scheme gives a physically reasonable solution.3(B). Two implementation strategies of the corrective scheme.16 has a flow with 8 = 63. The full correction of the crossflow diffusion gives the true solution without any numerical diffusion error. The mesh point implementation gives an unphysical solution . 5. Fig. mesh point and mesh interface.20 shows the diffusion corrections of the mesh point and mesh interface implementations at each interface. although not identical to the true solution.15 numerical solutions are given for a simple pure convection problem with various diffusion corrections. The flow is at a 45* angle with respect to the mesh orientation. 5. Both of them are tested in a recirculating flow problem with pure convection.430.141 constant from the total diffusion constant of the finite difference equation. Figure 5.21 gives the true solution. donor cell solution and the solutions by mesh point aand mesh interface implementations. 5. Figure 5. Figure. are introduced in Chapter 4. The corrective scheme is also tested in the problem geometry of Fig. 5. In Fig. 5. Fig.17 and Fig.
29 5.75 6. 0 V ________ D=0.01 7.88 5 4.0 2.0 2.88 10 5 0 2.12 5.71 0.29 0.0 10 7. Donor cell solutions with various diffusion corrections for a pure convection problem with e = 450 .5  3. 5.41 u=v=l Ax=Ay=l D =D =0.0 2.99 D=0.5 (True Solution) 9.5 0 1.142 D=0.5 x y Fig.92 5.0 1.71 5.3 D=0.08 5.59 7.0 8.15.25 0 5.13 7.1 10 8.
16.73 10 7. v=2 x= y=l 2 D =D = x y 3 Fig.33 1.430 . True solution and donor cell solutions with and without diffusion correction for a pure convection problem with e = 63.82 0.07 0.11 True Solution V 63.96 3.10 9.59 1.07 2.11 6.11 0.93 10 5. 5.91 0.37 3.143 D=0.41 0. 0 D2 3 0.04 4.07 10 3.56  2.430 u=l.
17.144 Analytic solution Numerical solution with donor cell Solution by corrective scheme 10 A spatial coordinate Fig. 5. 5.3(B) . Comparison of the analytic solution. numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig.
5.3(B) . 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.145 Analytic solution Numerical solution with donor Solution by corrective scheme 10 5 0 spatial coordinate Fig. 5.18.
146 0 Ofr 3 ~LC~' / 10 1 1 5 0 1 3 3. 5. 3 ) 2 3 2 Lx = Ay = 1 Fig. Simple recirculating flow field for test of the implementation strategies of the corrective scheme .19.
5.75 0.20.75 1.5 0.147 0.125 0.75 1.25 1.2 0.19 .25 mesh point implementation 0.75 0 0 0. 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.2 0 0 0 mesh interface implementation Fig. 5.5 0 0 5 0 0 .625 0.75 1.625 0 .
10 9.83 6.21 mesh interface implementation 0.19 .75 6.08 9.21 7.82 8. Comparison of the true solution. 5.16 11. 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.94 9.21.26 8.148 True solution Donor cell mesh point implementation 5.86 12.18 7.51 10 5 0 Fig. 5.10 8.16 6.
28 show the steam concentration distributions after one time step. 5. After a while a flow field is set up in the This containment by the input mass and momentum of steam. The steam concentration distribution at the beginning is given in Fig. Since the steam concentration distribution and flow field are given at the start of the calculation.25 and Fig. respectively. 5. The true solution cannot be reproduced because some of the information about the flow field has already been lost in Fig. 5.149 while the mesh interface implementation gives a physically reasonable solution. the steam concentration for the new time step can be calculated in an explicit way. 5.26 show the diffusion constant corrections of the mesh point and mesh interface implementations and Fig.3 sec with the corrections in Fig.23.24 in which the source mesh has the highest steam concentration.28 is not a reasonable solution while the mesh interface implementation result in Fig. The result in Fig. There was initially air in the containment and steam is introduced in the source mesh at the rate of 0. 5.22 shows the flow field in the containment and more detailed information about the flow field is given in Fig. Figure 5. Another test calculation is done for a two dimensional.25 and Fig.19. 5.2 kg/sec. Figure 5.26. 5. 5. 5. 0. flow field is used here to test the mesh point and mesh interface implementations of the corrective scheme.27 and Fig. 5.27 is physically reasonable and even better . rectangular containment with an arbitrary recirculating flow field.
150 Ax = 0. 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 = 1.0m Fig.22.
2474 04 1379 0.0937 0 1.1385 0.3074 0.93 2.2216 0. 5. 861 1. 1317 0.0348 0.0127 0.4920 0 .1139 0 .1422 2148 0. 3563 0.3 095 1.2460 0.2190 0.5174 1.151 0.531 0.34 0.3957 0. 045 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 .3253 0.2159 0.5 m Ay = 1 m Fig.4870 0.2629 0.1276 0 .2435 0.5893 0.1229 0. 5955 SOURCE 0.207 2. 2214 velocity [m/sec] Ax = 0.0492 0 .7126 0. 0735 0.23. 0648 1.456 0.
1859 SOURCE 0.1266 0.0819 0.1750 0.0298 0.0031 Ax = 0.0358 0.1178 0.0305 0.152 0.0136 0.1085 0.0095 0.1244 0.0049 0.0620 0.0002 0.1420 0.1592 0.0007 0.5m by = Im Source: steam concentration in Ikg/m 3 0.24.0346 0.1710 0. Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .0676 0. 5.0179 0.1712 0.0011 0.2 kg/sec of steam Fig.
C209 0.0224 0. 0.0 183 Ax = 0.25. 0. 0488 0.0 278 0.0794 0.stant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment . 2422 0.0937 0.1279 0.0590 0.2740 0.5m by = Im Diffusion constant in Im2/sec] Fig.1451 0. 0396 0.0348 0.0300 0.1694 3106 0. 0152 0.1452 0.0311 0.0731 116 0. Diffusion co.0193 0.153 0.1178 0.1763 0.0253 0.1014 258 0. 0.0311 0.1 389 0.1583 0.3262 0.0 362 0.0600 0..0255 SOURCE 0. 5.1701 0.1854 0.0440 0. 0227 0. 0086 0.
154 0 0 0.4732 0.0479 0.1026 0 SOURCE 0 0 0.0 76 0.2134 0.0860 0.1149 0 0 0.0443 0.2529 0.9769 0 0 0. 0706 0. 5.5m by = lm Diffusion constant in Im2/sec] Fig.0632 0.0215 0.0 923 0 0.0 257 0.26.3694 0. 0240 0. 0579 0.2317 0 0.0 111 0 Lx = 0.0518 0 0 0.0534 0.0192 0 0.0287 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 .
0735 0. 5.0833 0.1733 0.27.1243 0.0342 0.1847 SOURCE 0.155 0.0107 0.1475 0. with donor cell differencing convection term without any diffusion of correction .0003 0.3 sec.0046 Lx = 0.1623 0.0448 0.1363 0.0011 0.0195 0.0068 0.5m Ly = Im steam concentration in [kg/m 3 ] Source: o.0342 0.1135 0.2 kg/sec of steam Fig. 0.1321 0. Steam concentration distribution after one time step.0016 0.0465 0.0947 0.1706 0.0233 0.1711 0.
1747 0.0005 0.156 0.1797 0.0029 0. Steam concentration distribution after one time step.0433 0.0465 0.5m Ly = Im steam concentration in Ikg/m 0.28.3 sec.0287 0. 5. 0.0355 0.1729 0.0092 0.0250 0.1646 0.1911 0.0709 0.0068 0.0937 0.0042 Lx = 0.1493 0.2 kg/sec of steam Source: 3 Fig.2086 SOURCE 0. with mesh point implementation of the corrective scheme .0229 0.1247 0.1334 0.0010 0.1240 0.0815 0.
1112 0.29.157 0.0360 0.1756 0.1310 0.0424 0.0228 0.0722 0.0815 0. with mesh interface implementation of the corrective scheme .0058 0.1847 SOURCE 0.0007 0. 5.0039 ax = 0.0003 0.5m by = Im steam concentration in [kg/m 3 Source: 0.1738 0.3 sec.0100 0.0192 0. 0.1243 0. Steam concercration distribution after one time step.2 kg/sec of steam Fig.0010 0.1342 0.0947 0.0481 0.0295 0.1487 0.1628 0.1794 0.
In Fig. =10.31 shows the profiles of 4 along the dotted line for the time step size of 0.0. 5. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable. 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. when the Courant limit is 1 sec. The results of the ADI and explicit schemes are compared in the problem geometry of Fig.30 there is an inflow from the bottom surface at the velocity of v=1. In the meshes at the top of the source mesh. 5. which is exactly the .0 and the inlet boundary value of 4 is.32 shows the profiles of 4 at the same loca tion for the time step size of 1 sec.158 than the donor cell solution. 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. 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. it has the maximum time step size that can give a physically reasonable solution.30. Figure 5. The ADI and explicit schemes give almost identical results during all the transient. Figure 5.3 ADI Solution The ADI scheme is tested to use a time step size larger than the Courant limit. 5.5 sec.
v = 1. 5.YXI~^.1I I _~ 1_.30.0 c= 1.~ ~ILI_~_~LI*L~l*IIIU*XI 1 9 IB Source of ' 'A Profiles of 4 = 10.0 on this line are compared. Problem geometry for comparison of the explicit and ADI solution schemes .0 Fig.0 Ax = Ly = 1.
0 11.160 12. 5.01 12.0 10.0 10.0 12. 0A Explicit ADI 10.0 11.0 12.5 .30 the profile of sec and the Courant limit of 1.31.0A 12.0 11.0 10.0 0.0 11.0 sec for Lt=0.5 sec 12. 5.0 coordinate 11.0 A iatlai B 10.0 10.0 11. Comparison of the explicit and ADI solutions for along the line AB in Fig.0 Fig.
0 Fig. 12.0 ~ spatial coordinate 12.32.0 16 sec 20 sec 12.0 sec and the Courant limit of 1. 11.0 11.0 sec .0 10.30 for At=1.0 11.0 11. 5. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig.161 1 sec 12.0 10.0 10. 5.0 Explicit ADI 10.0 4 sec 11.0 10.0 .0 12.0 12.
No crossflow diffusion corrections are involved in the results of 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.162 the Courant limit.31. 5. Initially.33. 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. 5. 5. there is some deviation between the two results. 5.32 and Fig. Fig.
Comparison of the ADI solutions for the profile of at t=10.0 At =0.33.5 sec ec 11.30 with different time step sizes . 5.0 sec along the line AB in Fig.163 12.0 t spatial coordinate B Fig. 5.
2 Numerical Schemes (1) There are two numerical diffusion sources. (2) The models of laminar and turbulent diffusions in Chapter 3 may be adequate for predicting the hydrogen transport in the containment. The total diffusion con stant is the sum of the laminar and turbulent diffusion constants.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.164 CHAPTER 6 CONCLUSION 6. The thermal equilibrium is assumed and the phase change occurs to maintain 100% relative humidity. The mass diffusion. The con tinuity/momentum equations are decoupled from the scalar transport equations and solved by the SMAC scheme to obtain the flow field. energy. Crossflow diffusion occurs due to the donor cell treatment of the convection term in a multidimensional problem. or superheated steam as may be appropriate. and turbulence equations are solved separately using that flow field. 6. The . truncation error and crossflow diffusion. in the finite difference donor cell treatment of convection.
(2) Two types of approaches. have been tried in order to eliminate the numerical diffusion. The skew differencing scheme gives good results for some problems. ~unx. but not always.165 effective diffusion constants of the truncation error and crossflow diffusion are of the same order of magnitude. skew differencing and corrective schemes. It gives unphysical results for most recirculating flow and coarse mesh problems. 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 inclusion of corner points complicates the matrix structure and a fully implicit scheme should be used for maintenance of stability. recirculating flow problem is due to the crossflow diffusion. the corrective scheme is generally preferred to the skew differencing scheme. The corrective scheme is based on the fact that the additional crossflow diffusion can be predicted theoretically for every mesh at every time step. Therefore. convection dominant. . most of the numerical diffusion error in a multidimensional. Therefore. The truncation error diffusion occurs in the flow direction while the crossflow diffusion occurs in the diagonal direction of neighboring mesh points. It is conservative and an explicit scheme can be used without affecting the simple solution structure.
diffusion and cell Reynolds numbers. the physical constraint in section 4.4. The graphical method is used to obtain the stability condition in terms of the Courant. Cx. (5) The maximum time step size is limited by the Courant The Alternate Direction condition in an explicit scheme.. i. Implicit (ADI) scheme can be used to increase the time step size and decrease the overall computational effort. are tried for the corrective scheme. are identical in a unidirectional flow. using the characteristics of a quadratic equation. dx and Rx .e. The mesh interface implementation is always preferred to the mesh point implementation in a recirculating flow problem. mesh point and They mesh interface.166 (3) Two implementation strategies. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable. 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. (4) The Von Neumann stability analysis is applied to various finite difference forms of a general conservation equation. First. The results of numerical experiments are found to be consistent with the Von Neumann analysis.2 imposes a maximum time step size that can give a physically . it has its own limitations due to the following factors.
a steadyThe state solution may never be reached due to the fact that one time step is composed of a few fractional steps.167 reasonable solution. For example. Second. asymmetry of the given problem may give different results according to the sweeping sequence. the ADI scheme is a fractional step method. .
. etc. it is necessary to predict the flow regime in order to use appropriate physical models. The flow regime criterion may be replaced with a model that covers all the turbulent. time step size. multidimensional recirculating flow problem. transitional and laminar flow regimes. skewdifferencing and corrective schemes. A criterion for the flow regime should be developed in such a form that it may be implemented into a computer code. distance from the wall). previous time step information. The model should show proper limiting behaviors as the Reynolds number goes to infinity (turbulent) and zero (laminar).g. . The important parameters may be the velocity. to eliminate the crossflow diffusion. velocity gradient. 7. They may still be improved further to get an accurate solution in a transient. geometry (e.168 CHAPTER 7 RECOMMENDATIONS FOR FUTURE WORK 7. The modified kE model in a low Reynolds number flow may be a guide in this approach.1 Physical Models Since both the laminar and turbulent flow regimes are possible in a complicated geometry and flow field.2 Numerical Schemes There have been two approaches.
The stability of the skewdifferencing scheme is open to question in the explicit. . Skew differencing scheme 1. More tests and validation calculations are required in recirculating flow problems. ADI and implicit schemes although only the implicit scheme has been used up to now. 2. The solution in Appendix A may be a useful guide in this work.169 Corrective scheme The corrective scheme has been successful by the mesh interface implementation strategy in the scope of this work. It may be possible to develop a reasonably simple form that may or may not include the corner points. The conservative form of the skew differencing scheme is complicated and time consuming in comparison with the corrective scheme.
R. and Lightfoot. F." Journal of Comp.. R. H.. and Patankar. 3.N. Boyle. 372397 (1977).R.H. "Transient Effects in Turbulence Modelling. thesis (1979). T... E. Briggs. C.. Briley.B.H." LA4370 (May 1970). A.." Journal of Comp.V.User's Manual.D. Bloom." MIT Ph. 5473 (1980). 393409 (1980). Engineering Calculation Methods for Turbulent Flow.." HEDLSA2682 (October 1982)." Numerical Heat Transfer 3..G. Methods in Applied Mech.R. "Solution of the Multidimensional Compressible NavierStokes Equations by a Generalized Implicit Method. Broadus. D. and McDonald. Phys." Comp. Cebeci. 6. 24. Transport Phenomena. and Eng. Comp. "BEACON/MOD 3: A Computer Program for Thermalhydraulic Analysis of Nuclear Reactor Containments . (April 1980). and McDonald. J.R. Bird. F. A. Baliga. and Whitelaw. Briley. Phys.R. P... "A Finite Difference Scheme for the Incompressible AdvectionDiffusion Equation. 233241 (1975).." Journal of (1968). W." NUREG/CR1148 (EGG208). Phys. 6. "A New Finite Element Formulation for ConvectionDiffusion Problems. W. B.H. John Wiley & Sons (1960).A. 94110 Amsden. Stewart.. et al.R.. G. 332325 (1970). Bradshaw. et al. D. A. "On the Structure and Use of Linearized Block Implicit Schemes.170 REFERENCES [1] Amsden.H. "The SMAC method: A numerical technique for calculating incompressible flows." Journal of Comp. Academic Press (1981). and Harlow. "Hydrogen Distribution in a Containment With a HighVelocity HydrogenSteam Source.A. F. H. "A Simplified MAC Technique for Incompressible Fluid Flow Calculations. 34. [2] [3) [4] [5) [6] [7] [8) [9) [10) [11) [12] . W.. Phys. and Harlow. "Transport of Turbulence in Numerical Fluid Dynamics. Amsden.A. S.E. and Harlow.
Crowley. J. C. "Comparative Analysis of Numerical Methods for the Solution of NavierSTokes Equations... G. Chen.. Keller. Nielsen. 1976). 1 (Nov.D. and Romero.B.. Cloutman. "Numerical Methods of Reactor Analysis. 471484 (1967). J. P.K.A Computer Program for Calculating Timedependent Turbulent Fluid Flows with Slight Density Variation. 127137 (1982). and Design 69.W. Cook. and Nakayama.G. Jr. M. and Lewis. Covelli. Vol. Academic Press (1964)." MIT M. [14] [15) [16] [17] 118] [19] [20) [21] [22] 123] [24] [25) . J. Methods in Applied Mech. thesis (1981)...I." Nuclear Eng. Phys." MIT Ph.S. Hirt... and Mallison. "Studies of Numerical Methods for the Plane NavierStokes Equations. R.F.D. thesis (1980). and Eng.D. G.H. N. "Accuracy of Difference Formulation of NavierStokes Equations." Journal of Comp. 2943 (1976).P. "SecondOrder Numerical Advection.J. G.B.C. 27. S. "VARR2 ." The Phys. K. Clark... De Vahl Davis.. J." Nuclear Eng.D.. "Buoyancy Effects in Mathematical and Numerical Turbulence Modelling. L. "Coolant Mixing in the LMFBR Outlet Plenum. H. T. Phys. S. and Design 69. M. 32. L.S. "Tensor Viscosity Dukowicz. Method for Convection in Numerical Fluid Dynamics. "Simulation of Containment Cooling with Outside Spray After a Core Meltdown. Williams. and Hansen. Varadi. Hirsh... "An Evaluation of Upwind and Central Difference Approximations by a Study of Recircul ing Flow. 1344 (1981). of Fluids Supplement 2 (1969)." LA6236 (July 1976).. and Ramshaw." MIT Ph.. Chang. S." Comp.D." CRBRPARD0106. "Study of Nonhomogeneous Conditions in Reactor Containments Due to Separation Effects. Vol.171 [13) Carbajo. 95107 (1982). thesis (1977). Y.." Journal of Comp. 7179 (1979). P. 1. Cebeci. B.. Chang.H.I... "SOLAICE: A Numerical Solution Algorithm for Transient Compressible Fluid Flows.L." Computers and Fluids. Cheng. W. 4.
The 1978 Freeman Scholar Lecture. R." The Phys. M." Journal of Comp. 339355 (1968)." The Phys.. (1975)..F.O. G... and Marohar. Hirt.. "A Comparison of Hybrid and Quadratic Upstream Differencing in High Reynolds Number Elliptic Flows. Turbulence. and Amsden. 309312 (March 1975).E. Harlow. 17.. R. R. Harlow.H. J. "Improved Form of the Low Reynolds Number kc Turbulence Model. and Eng." Journal of Fluids Eng. Phys.W. 557573 (1980).. 8. "Numerical Calculation of Multiphase Fluid Flow.A. of Fluids Supplement 2 (1969)... Hamza. [38) ." MITEL 81020 (June 1981). F. [27] [28] [29) [30) [31] [32) [33) [34] [35] [36) [37) Hirt. New York (1954)." Journal of Comp. Phys. of Fluids. Hinze.A. C.A. J. Humphrey. 2.E. 101. C. 8195 (1981). 29. of Fluids Supplement 2 (1969). "A Numerical Fluid Dynamics Calculation Method for All Flow Speeds.A. Vol..W.. Han. Vol..172 [26] Fromm. Gupta.. A.C.. T. F.. 15." The Phys. Vol. Hoffman. 1952 (1975).B.M.. B.W. for Num. Molecular Theory of Gases and Liquids. Methods in Eng.. 528 (March 1979).H. Harlow.H.. "Practical Investigation of Convective Difference Approximations of Reduced Dispersion. 3. and Amsden. "Computer Studies of Timedependent Turbulent Flows. A. 8093 (1968). "Buoyancy Induced Fluid Motions Characteristic of Applications in Technology . M. and Amsden. McGrawHill Book Company. Phys. J. and Bird. C. "Numerical Calculation of Almost Incompressible Flow. Hirschfelder.P. and Golay. A. Wiley.." Comp." Journal of Comp. and Launder." Journal of Comp.H. F. B... 197213 (1971). "Behavior of Buoyant Moist Plumes in Turbulent Atmospheres." Int. 2nd ed. Methods in Applied Mech. "On the Use of Central Difference Scheme for NavierStokes Equations. Curtiss. J. 18. "Heuristic Stability Theory for Finite Difference Equations.O. J. Gebhart. Phys..
Steam Tables. Phys. M. B. Lam.. W.B.. Vol. 269289 (1974). 23.P." Trans. "ADI Solution of Free Convection in a Variable Viscosity Fluid. and Spalding.J. Vol. Launder.A. Manno. "Practical Evaluation of Three Finite Difference Schemes for the Computation of SteadyState Recirculating Flows. Vol." Int.E.. et al. 3.D. Cl..G. B. 16.K.. 1981). "A Modified Form of the kc Model for Predicting Wall Turbulence. translated (October 1980). B. V.." MIT Sc. Heat Mass Transfer.P. Methods in Applied Mech." Int..G. Jones. "The Prediction of Laminarization with a Twoequation Model of Turbulence. John Wiley & Sons (1978). 15. M. and Keyes. 68. "The Calculation of Low Reynolds Number Phenomena with a Twoequation Model of Turbulence. M.. Manno.." Journal of Comp. "Experimental Investigation of Hydrogen Distribution in a Containment of a Light Water Reactor Following a Coolant Loss Accident. 16. part 3. J. and Golay. Jones.. 293312 (1980)..3633." BF63. V.E. Langer. J.. 456460 (Sept. and Eng.E.V. G. of the ASME. Heat Mass Transfer. thesis (1983). [40) [41) [42) [43] [44) [45] [46] [47] Leschziner. J. Frankfurt.H." J. Reece." Comp." Comp. "The Numerical Comparison of Turbulent Flows. 11191130 (1973). J. G. 221239 (1974). K. Jr. and Launder. Methods in Applied Mech. and Rodi. F.H. 103.. Launder. Keenan. Fluid Mech..173 [39) Houston.P." Report submitted to 2nd International Topical Meeting on Nuclear Reactor Thermalhydraulics (March 1982). and De Braemaecker.W. 537566 (1975). and Launder. B.. D. 148) [49) . W.. "Analytical Modelling of Hydrogen Transport In Reactor Containments. and Bremhorst.P. "Progress in the Development of a Reynolds Stress Turbulence Closure.E. and Eng. 301314 (1972). Battelle Institute e. C. W.. Vol. "Analytical Modelling of Hydrogen Transport in Reactor Containments.
and Spalding. and Sherwood..D.. Methods in Applied Mech. 9.D. and'Eng. Phys. Hemisphere Publishing Corporation (1980).J. Roache. J." Computers and Fluids. Raithby.B. Heat Mass Transfer. 386389 (Sept. R.D.174 [50] [51) Nakamura.J.. Roache. Hermosa Publishers. Vol. Inc. W. K.. 541550 (1972). S. Roberts. "A Note on the Empirical Constant in the KolmogorovPrandtl EddyViscosity Expression. Rodi. 15. "A Critical Evaluation of Upstream Differencing Applied to Problems Involving Fluid Flow." The Phys. "On Artificial Viscosity." Mathematics of Computation. New Mexico (1976)..V.E. "A Calculational Procedure for Heat.... Methods in Applied Mech. Computational Fluid Dynamics.. K. P. Raithby.. G. Patankar. S. 9. 153164 (1976). Raithby. "Convergence and Accuracy of Three Finite Difference Schemes for a Two Dimensional Conduction and Convection Problem. Turbulent Buoyant Jets and Plumes. "SkewUpstream Differencing Schemes for Problems Involving Fluid Flow." Int. 1975). The Properties of Gases and Liquids.K. "Numerical Experiments with a Difference Model for the NavierStokes Equations (Turbulence Model)... Pergamon Press (1982). Numerical Heat Transfer and Fluid Flow. (1977)." Journal of Comp. 2. Computational Methods in Engineering and Science. "Upstreamweighted Differencing Schemes and Their Application to Elliptic Problems Involving Fluid Flow. Albuquerque. 10.. Methods in Eng.C... [52) [53) [54) [55) [56) [57) [58) [59] [60) [61) [62] [63) .B. 4. S. for Num. N. 75103 (1976).. and Eng. G. Runchal. G." Comp. Reid. W.V." Comp." Int. and Weiss. Schonauer. 169184 (1972)." Trans. Patankar. T.. of Fluids Supplement 2 (1969). D. of the ASME. 17871806 (1972). McGrawHill. P. Mass and Momentum Transfer in Three Dimensional Parabolic Flows.O. John Wiley' & Sons. 20 (1966). J. and Torrance. New York (1958). 191206 (1974). W.. A. "Convective Difference Schemes. Rodi.K.
"Development and Validation of a TwoVariable Turbulence Model. "Diffusional Properties of Multicomponent Gases. and Valentin.175 [64] Shin." Numerical Heat Transfer..B. . and Strong..A. and Woolner.. Smith..L. 3." of the National Bureau of Science . J. A.. Vol.W. M.. A. R. and Hutton.. No.B. 20. 220237 (1976).D." Comp. Spalding. A.Dec. J. 2... G..E.. Phys. Y. G. (Jan. G.S.L. 4 (Oct. and Eyler. H. 95102 (1950). MIT Press (1970). 551559 (1972). Budden. 411428 (1980).B. R. 35. bulence. and Eng... L. J. "Comparison of Finite Difference Journal of Computations of Natural Convection. "The Numerical Treatment of Advection: A Performance Comparison of Current Methods." Journal of Comp. For Num.M." Int..A.D. 411428 (1980). D. G. Methods in Applied Mech. Wilke. 4. Raithby.D.D. "Simulation of Convection and Diffusion Processes by Standard Finite Difference Schemes and by Influence Schemes.R. Methods in Eng. Tennekes. 46." Chemical Engineering Progress..H. Vol.J." FATE80114 [73] (Jan. Trent. D. 1974). "Proposal for a New Discrete Method Based on an Assessment of Discretization Errors.. C. No. 3.B. 153168 (1982). and Lumley. 72B. "TEMPEST: A Threedimensional Timedependent Computer Program [72] for Hydrothermal Analysis. 1968). MatheResearch matical Sciences. Stuhmiller." Numerical Heat Transfer. "A Novel Finite Difference Formulation for Differential Expressions Involving Both First and Second Derivatives. K. A First Course in Tur [65] [66) [67] [68) [69) [70] [71] Torrance.G. K." SAI74509LJ. "Numerical Analysis of Fluid Hammer Waves by the Method of Characteristics.. 1981). Strong. Raithby. Stubley. Stubley.
2 is simplified further by substitution of 4 for T as follows.1.176 APPENDIX A ANALYTICAL SOLUTION FOR THE PROBLEM IN FIGURE 5.3) ax .1) In Eq. 2 x 2 a2 ay = c 2 (A. 5. c. pc u aT = k(a2T + 2aT 2 x2 p x ax ay (A.3(B) and the constants can be grouped into one constant.2) Equation A.1 is2 reduced to the following form.3(B) A steady state two dimensional energy conservation equation without any source is given in the follbwing Eq..1 convection occurs in the x direction and diffusion occurs in both x and y directions. A. a= k PCp C =2 u 2A. as follows. Therefore. A.a2T aT 2 + a2 2 2c ax ax 2 y ay (A. Eq. The boundary conditions are specified in Fig.
which are functions of x and y only respec (x. A. The variable 1 is separated into two vari X and Y. A.177 where 4 = eCXT (A.8) dy where a > c > 0 .3 can be solved by the method of separation of variables. A.6) dx dy The two terms on the left hand side of Eq.a =  (A.4) The boundary conditions should also be expressed in terms of 4 as given in Fig.3 is reduced to the following form. Equation A. I d2X + 1 d 2Y 1 = c2 C (A. tively. ables. c 2 X cx 2 = (A.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.7) d Y c .1.y) = X(x)Y(y) (A.5) Then Eq.
Boundary conditions in terms of 4 for the problem in Fig. 5. 4 = ec T . A.= YO T 0 ii ii i h YO + co = 0 4 = 0.3(B) where 0 is given by.0 0 = 0 x Fig.178 .1.
9 and Eq. A. 13. The boundary con dition at x=0 will be considered later.14) (A. A.10 should satisfy the following relations. Eq. dY dy y=O Y=O 0 (A. A.7 and Eq.16) .8 can be easily obtained as follows. A.15) ny 0 = n C2 _ n+c 2 e nX 1 n (a n 9 C) (A. A. X = Cleax + C2 eax (A. A.9) Y = C3sin8y + C4cosSy (A.11) dY dy=Y 0 = 0 (A.12 and Eq.1 may be given in terms of the variables X and Y as follows. 13) In order to satisfy the boundary conditions given in Eq. the constants in Eq.10) The boundary conditions in Fig.11.12) (d dx _cX) = 0 x=xO (A.179 The solutions to Eq. A. C3 = 0 (A.
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. nw sin2 nO0 (n=1.18) n n Since the boundary condition at x=0 should also be satisfied. Therefore.0 2 A n Th h nr 1 n 1+( a c )e n 1 a +c 2a x .19) . the constants A.1 are determined as follows. .2. Th 0 2. 17) T(x.y) where nr (A.y) = e cx (x.180 where 2 =c 2 + n (O) n 2 yO =c 2 + 8 2 .3 .) (A. . the final solution can be expressed as follows.
i.4. When the tilde phase of the SMAC scheme and other scalar transport equations are solved by the ADI 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. Equations B.iillaY~i~~l_ ^YII^^~II ( .1 and B. B. there are six sweeps in a three dimensional problem and four sweeps in a two dimensional problem. The most important sweep should be given the first priority and most updated information. 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.1Eq. 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. the computational effort for one time step is greater than that of the explicit scheme.3 are the z and x direction sweeps of the x direction momentum . B. The ADI scheme is applied to a two dimeraional momentum equation in the following Eq. They are formulated for a natural convection problem where the gravity acts in the zdirection. The sequence of the sweeps is arbitrary and depends on the problem under consideration. Therefore.
u Lr * n+1 . u n * u j i+ Ui+D Dt j + * n i+ j+ Ui+ j .182 equation.2 and B.Rxi+ jui+ jui+ j RX * n+l (B.4 are the x and z direction The z direction sweeps of the zdirection momentum equation.2) n+l i+j * .Wi+ Az j Ui+ j 1 ui+ j+l * .3) .2ui+. ar .2ii+ Ar 2 + Wilj+ .u n+l ui+2 2 n+l Ar2 +2u j n+l ui j . 2 Az + Ui+ 1 P  n  i+j i+ jui+4j (B. sweep of the z direction momentum equation is treated as the last one in order to be given the most updated information. Equations B.P z  n * RZij+ ij+ ij+ (B.1) * * n * *  ij+ At Wi+1 j+ ij+ + i+Ij+wij+ uij+Wilj+ 1 P n~ .ui+ j At  n+l * u i+lj i+ .
183 n+l i+ * wij+ * i n+l +lWij+ * n+l zW.4 have given almost identical results with the explicit scheme for a time step size below the Courant limit. w At n+l Wi.4) Equations B. B.3. and B.1.2. . B.+ n+1 n+l Az 2wi + w _P z P0 zP g z RZ n+l ij + wij+ 3wij+ (B.
'difprg fortran'. is the program for testing Fig.184 APPENDIX C COMPUTER PROGRAMS The files. There are When there are inflow and outflow. the calculational logic of the mesh interface implementation of the corrective scheme in section 4. 5. When there are two outflows in any of the x.2. or all inflows in x.3(A).3(B) and Fig. 5. two different cases treated separately. the crossflow diffusion constants can be calculated in a simple way as in the first part of the program. are assumed to be zero. 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. 'progl fortran'. 'prog2 fortran' and 'prog3 fortran'.. The file. Fig. are the computer programs for calculating the numerical values of the analytical solution in the problem geometries.3.8. y and z directions. 5. y and z direction mesh interfaces. All the inflow velocities The component velocities are used to calculate the crossflow diffusion constants for eight subspaces in a three dimensional coordinate space.8. 4. a flow split occurs as shown in Fig. Then four crossflow diffusion constants at every interface are summed up to give a total crossflow diffusion constant .4.
. It is used to solve a one dimensional implicit finite difference equation in the ADI scheme. The total crossflow diffusion constant can be calculated in the same way for every mesh interface. has four neighboring subspaces. The ADI scheme for the momentum equation is given in Appendix B.185 because one axis direction. The subroutine 'tdm' solves tridiagonal matrix problems by forward and backward sweeps. A portion of the VARR program is introduced to show how the ADI scheme is implemented in the energy and momentum equations. the positive x direction.
ALP II FOIRMAT(F7.NY.COF*DSIN(II*HPI)/(1.O) ALNX .215.04ALP) CSO .1 ) DX.I=l.000190 PI00O200 NY*OX YSO v PI*PI/(YO.N FILE: PROG FORrRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT REAL*0 (AH.14 55 ATOT * THIODEXP(ALN*XO)/2.4ALNX) ATOT .YO) CCOrrF 2.O*ALNXO)/(ALNC) COF * CCOFF/lI AAN .Tl.0 00 20 IPsi.U.0 IHPI DO SUM 10 JJI. AAN*2.1415926536 C = U/(2.2) PI a 3.5F11.OALNOEXP(ALN.EO.0 44 YY I IIPI*(UJ0.XO)/(ALNC) GO TO .NY  0.0*TII/PI PI/2.5)/NY ADD * ATOTTDCOS(YY) SUM T(d) v SUM + * ADO * DEXP(C*XO) 20 CONTINUE SUM 10 CONTINUE WRITE(G.3) SlOP END PRO00210 PRO00220 PROO0230 P1ZOOO240 P1I000250 PROOO260 PROOO270 PROOO200 PROOO290 P11000300 PROO0310 PRO00320 PRO00330 PROOO00340 PRO00350 PROOO360 PROOO370 PROOO300 PROOO390 PROOO400 . C*C XO " NXOX YO * PROOOO0 PROOOO20 PROOO30 PROO0040 PRO000OOOO P14000070 PRO000OGO P11000090 POOOOT100 PROOOO10 PIIOO0120 PRO00130 PROOO140 PROOO150 PRO00160 PROO000170 PRO00100 P1.3F7.100) (T(I).2.IALNIC)*DEXP(2.NY) 100 FOnMAT(IOX.26 II IF  IP  I GO TO 55 ALN = DSORT(CSO*II*II*YSO) (II.NX.OZ) T(tO) DIMENSION REAO(10.
3F7 .0Z) DIMENSION TII(10).U/(2.2.21 II = IP  I ALN = DSORT(CSO + IIII*YSQ) EXALO = DEXP(ALN*XOO.5) EXAL = OEXP(ALNOX) XAL = DEXP(ALN'X) EXALN D EXP(ALN*XN) XALN = DEXP(ALNXN) IF(II.5+YO 00 20 IPs1. NX.0 SUM3X 0. TL(10). TC(IO) TH14.DCOS(BEN*YC) .0 = YO/2.XN = XN . GH(10).OALP) CSO = C'C XO * PROOOO O PRO00OO20 P R000030 PROOOO40 PROO00SO POOG0050 PROOOO000 PROOOOOO PROO0090 PIR000 100 P11000110 PRO00 120 PROOO 1.30 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 = 2.2) PI .COF*DSIN(II*HIPI)/(I.OCOS(OEN*Y4) Y4S DOSIN(3EN'Y4) ADO * AAN*(EXALO + ALNX*XALO)*YOO PROOO520 PR000530 P1OOO00540 PR000550 . ALP READ(10.0 X + XO/2.1.0 SUMO SUMI r 0.0 SjU3 " = 0.215.0 X = (JJO. NY. 11 FORMAT(F7.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN .5) XALO = DEXP(ALN*XO*O.(ALN#C)*DEXP(2.O.3.0 = 0.O) GO TO 55 ALNX .0 .OTHIH/PI 11PI = PI/2.5*YO .DSIN(BEN+Y3) Y4C . GL(10). IC1l REALi0 (AII.+ALNX) BEN 11*PI/YO YOO .0  RTO DO =  OSORT(XXX) 10 JJ=1.X = 1.11) DX.0 NY2 XXX NY/2.5)*DX = YC = (JJ.DCOS(BEN*YI) YI Y22 • DCOS(BENY2) OCOS(BEN*Y3) Y3C Y3S .p A FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 MPI. U.0 .0 SUMaX SUM4 XN vI Y2 Y3 Y4 • 0.0 SUM2 = 0.21)*DX = X 4 YO/2. 1415926536 C .NY2 0.FO.
3) STOP ENO SUMO*DEXP(C*XO0O.(" % FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 ADOI w AAN*(EXAL+ALNX*XAL)*YII AD02 .EXALN + ALNX*(CALN)*XALN)*Y3C EN*AAN#(EXALN . 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.SFII.I 1OO FORMAT(IOX.tOO) (TL(l).NY2) WRITE(G6.0 AOOX a 0.5 A002 * TMIIIXAL0.II.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 .II.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(6.5 ADO3X 0.00) (GL(I).NY2) = t.NY2) WRITE(6.NY2) WRITE(.5 A)DO * TIHl'XALO*O.0 ADO3Y * 0.5) SUMI*DEXP(C*X) SUM2*DEXP(C'X) (SUM3X+SUM3Y)*DEXP(C*XN)/RTO (SUM4X+SUM4Y).II.AAN*(EXALALNX*XAL)*Y22 ADO3X v AAN*((C+ALN).I WRITE(6. 1OO) (TH(I).0 ADOvY * 0.NY2) = I.100) (GH(I).tOO) (TC(I).
NX.5)YO/(2.NY2 SUMO = 0.0 SUM3X = 0.(dJ0.5)*(XOYO/1r3)/NY2 YC=(JO.5)YO/(2 .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 . TL(IO).0)/NY2 X4XOYO/(2.0 SUM4X = 0.C.OtTIIIM/PI HIPI = PI/2.+ALNX) OEN = II'PI/YO YOO .5)*YO/(2.0) PROO000120 PROO000130 PROO0140 PROOO150 PR000O160 RT3 = C SORT(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) . FILE: PROG3 FORIRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT DIMENSION C 11 REALtO (AH. ( 10).DCOS(iEN*tY2) YV3C = DCOS(BEN'Y3) Y3S = DSIN(BENtY3) YAC = OCOS(3EN*Y.DCOS(UENOYC) YIl DCOS(OEN'YI) Y22 .0 SUM3Y = 0.0.0)/NY2 X2=(JJ0.5) * (YO/3. TC(1tO) PROO0020 PROO0030 PROO0040 PnOOO050 PROOG000 READ( 10. U.0*RT3#NY2) Y2=YO/2.2.22 II IP I PR00360 PRO00370 PROOO3UO P(1000390 P000400 ALN = DSORT(CSO + II*laIYSO) GO 10 55 Ir(II.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN = COF*OSIN(II'HPI)/(I.1415926536 C = U/(2.5)*(XO/RT3YO/6.SO5T(2.02) 111(10).0.0*NY2) X3YO/(2. I 1) DX.2) PI = 3.0 NY2 nRo P000O070 13100000 PROOOOOO PROOO0000 PRO00110 = NY/2.0(JJ0.215. TIIl.o) ALNX = (ALNwC)tDEXP(2.0 PR00O170 PR000100 PROO0190 PR000200 PlO00210 SUM2 = 0.RT3)+(dJO.3F7.O*RT3)+(JJO.XO/RT3.0 . NY.*0RT3#NY2) Y4=YO/3.0*RTJ))/NY2 YIYO/2.S)*(XO/RT3YO/6.NX*DEXP(ALNXI))*YI I ADD2AAN*(DEXP(AI.0 SUM4Y = 0.0 XC=Y0/(2. rORMAT(F7.0*RT3))/NY2 Y3=(JJO.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.EO.0) DO 10 00=1.O' XO/RT3)/NY2 XI=(dJ0.S)#(XOYO/(2. G1(10).0+(dJO.5)*(XOYO/(2. ALP PROOOO10 GL( O).0 SUMI = 0.5)*YO/(2.
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.X4) ALNX*(CALN)*OEXP(ALN*X4))*Y4C ADO4Y.NY2) WRITE(6.1OO) (TL(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.. 100) (GI(I).NY2) WRITE(6.I1.5 ADD2TtII#DEXP(ALN*X2)*O.100) (GL(I).100) (TH(I).NY2) tOo rORMAT(IOX.5 ADDH11H*DEXP(ALN*XI)*0.5+SUM3Y*0.OEN*AAN(DEXP(ALN*X4)4ALNX*)DEXP(ALN*X4))*Y4S GO TO 77 55 ADDOTIIII*DEXP( ALN*XC)*O.0 ADO3Y = 0.NY2) WRITE(G.0 ADDOX a 0.I= .5*RT3)*DEXP(COX3) GHi(JJ)(SUM4X#O.11.I=I.I=.NY2) WRITE(6.100) (TC(I).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 ADO3X PROOOSGO PROO00570 PR1000500 PROO000590 PROOOGOO P1100000.5)*DEXP(C*X4) 10 CONTINUE WRITE(6. PROOO6tIO .5*RT3+SUM4YtO.0 77 SUM1 .0 ADO4Y = 0.3) STOP ENO . = 0.
O.UU/DFLR PPZ * VV/OELZ PPT .2.EO.UM.OR.O.) ISA .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 .O.VELZ(0).DIFTA(0) REA( 10O.O.OR.5.0 RFGCR 0.GT.VP.DIFZ'IS3 * DIFZ*IS4 RFGr GO TO 00 AOS(UM) + ABS(UP) DI F00170 DIFOO IGO DIFOO00190 DIFOOlOO ISI IS2 IS3 IS4 155 ISG .ISt + IS2 ISO IS3 + IS4 ISC ISS5 + ISG C UI = 0.0 RFOT 2 0. bt ts .PPX $ PPZ + PPT GO TO 500 IF( PSUM.0 n I FOO 130 OMTI m 0.5*( VM V2 0.0) IF(UP.UU*DELR*( PPZ + PPT )/PSUM DIFZ * VV*DELZ*( PPX + PPT )/PSUM DIFT .WP 1SI * 0 152 .IFZA(0).0) IF(WM.EQ.EO.WW4DELTII#( PPX + PPZ )/PSUM C RiFGII .EO.2.O) IF(WP.UP.4 ~ FILE: O FI'IlG FOITIAN A VM/SP COINVERSATIONAL MONITOR SYSTEM PACE 001 DIMENSION VELX(0).L1.O.O.DEL'TII.DIFII'IS2 rGR RfGTI .DELZ.0 OMT .VELT(O).2) UU * UI + U2 VV .O 151 4 0 ISS1 O 156 O DIFOOOIO DIFOOO20 DIFOOO30 OIF0040 011:O0050 DIFOOOGO DIF00070 DIFOOOTO DIF00110 RFGRI = 0.0.LT.VM.0) IF(VP.DIrn*ISI .WW/DELTII PSUM .ISC.DIFRA(0)1.0.LT.0 IF(UM.O 15S3 .5*( VP WI * 0.5*( UP VI 0.t = 1 = i v 1 * .I 01o00200 DIF00210 DIF00220 DIFOO230 DIFOO240 DIF00250 DIF00260 DIFO0270 DIF00200 OlFOO290 DIFOO300 DI F003 10 I'I .0.GT.WM.( WM W2 0.VI + V2 WW = WI 4+ W2 PPX .IS .0 RFGTI .5*( WP C IF( ISA.GT.5) DELR.0 ) C DIFR .0) IF(VM.
EQ.0 oirZA(l) .) U2 4PF Z PET 1J2'( 1.0OPF Z) W2 *P FX' IF Z WI'.X(I)'DELR*( PPZ + PPT )0 'PSLJM DIFZA(1) m VELZ(I)*DELz#( rrX +.NE.0.( 1.( .0PFT) U2*PFZ*( 1.VEU. 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 *0.0.0PFX)#(t.0.PFX*PFT V2* I'F X'PF T V2.0.OPFX)O( I. oi rooG to Dl F00G20 D1F00630 01F00640 Ut 'PFZ .0PFZ) W20( 1.VELT(I)/OELTH PSUM a PPX + PVZ + PPT IE( PSUM.0PFT.PFT) WI#PI'X#PFZ WI'*ix( .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$( t.0PET) Vl*( 1.0PET) Vt .0PFZ) DIFO00650 oDI roofGo oir00670 nir006s0 mrof0069 ot rooioo DIF00710o VEIX(S) VELX(6) VIELX(7) VELX(0) VEL7( I) V El.o VELX( 1) VCL~X(7) VELX(3) VCLX(4 ) PFX i'rz PFT   U1/(UI'U2) Vt/(VI14V2) Wl/(W14W2) r DoIooseo 01F00590 DirooGoo.09'FX)#(t1.0PFX)#PFT V2( 1.PFT u i.0 oiETA(I) .0PFX)*PFZ DO 30 11.IPFZ)s( 1.0 0.FILE: DIFPRC'.0PFZ)'$( .0PFX)*PrET V2*( 1.o 'FX'( 1.0PFT) V t.VELX( I)/I)ELR PPZ .oPF7Z)'PFT UIs .0PFX)o( t. Z(2) vrI.0PFz) W2 &1EX'&( 1.O.( i.VELZ(I)/DELZ PPT .DELTH..NIE.NC.0PFX)'PFZ Wt'(I.0.0 (Vl*V2).V(LT(I).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 .PrT )/PSUM DIFTA(I) .0.0PFZ)*PFT U20( 1.*( PPX + PPZ )/PSUM GO TO 30 600 OIFRA(I) .0 PPX .0PFT) U I PFZ (I 1.0 ) GO TO 600 DIFRA(I) .0 Ir (W1#W2).0 Pr T Ir (UI+U2).0PET) Vi*PFX*( 1.
2) G FOIMAT( III .7) 5 FORMAT(9F5. 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 .2X.OMT GO TO 200 500 WRITE(G.2) 7 FORMAT( ItII .OMT I.RrGT I.20X.I.6F5.RFGR.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 .RFGT.
P(400).C .X(N).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.XN) DIMENSION A(N) .D.NMI .SUarIouT INE TDM(A .O(N).0.o(N).I KN a (X(I() X(K) RE TURN END  X(K41)*COK))/O(K) VAR09Z350 VAflo9:160 VA1109370 VA11093001 VAI.C(N).0t4O01 X( I ) I~) 13( ) 0(i1) NM I *N I 00) 2 1I.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 .
I vtR I5S0 rooOT 00G2 UC 'S(U(]K) ) VLRi 1210 vAR1E230 VAR E20 U(3KXX) ) LUGc .LE0.K) * 0.P2) GO TO 40002 (U(iK)*U(KX))) LUG = LS(O.0 D'rF2( I .TIONS SECT ION AD !ST ]F(ICLL].G. 1) GO TO 400 IF(K.E0.K) z 0.ItD.UG.EO.G  V RIS150 VAR.LT.CR.FFX ( .S .VGRD2/(.5(V( K)*:(.I.I.E1O 0005 FP" * IF(.K t KXX a jK . 1CONV..1) GO T0 49998 SOLUTiON FOR SIE.EO.JSP2 DO O40000 DO 4000 K.L!S( (]KZZ) ) .F2(J.0.4 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.KS 2.2D LUGRDX/(AUGOX +~DG'RD)' .LT.C .S( 4 . iB10 .C "hT.K) • AVG. DNF.DFFUSI'O =. D2)..PM..L. C C C C U AND v LDE VELOCITY EDU.E230 .P.0 OOC COI~'T2]'UE C VAR P182 vLR a210 VAR E220 VAR.K) a LUGDX.UG..S(0.xx'.NPC o1 ]F(K.K2NC JKZ2 .]K .O.. DA'' G.KBP2 " (3.KZ2))) .3. I O50 VI8090 K1 K12 2 KEP CONSTA'NT CLLCULATION C CRO5SFLD'.E0.)K2NC " (Ki)'NWPC .:X GO TO 40000 "0OO D.FX(3..) G0 TOrOOiC D'.
a uJD(2K) )0.PKKNUO(IPK) ).I) K2WC' + II)K NC *K2NC "00Z0100 K=2.I ) I KIPC # (KI)'tN'PCL VAR16D VAR 11510 VARi18520 !PK a 31. 054 (K.'O('.E.0.1 SRKSrI a * N BUOJY a LP0E7..I DIFFCD(3KL2) D!~FFCD(IKL3) *0(!.2)GO a TO 29990VALR iS.1. UD( 3K) UL UR UT Uo PK) UD( 3KP) U0( 3KMS) DCR DCL D!FFCD(IKL) D) F CO( KL.5 PC) '). 3F(3~.5 UODUMKS) 16D(1PK) * .0.E003 DO E0100O ) K GO TO50205 222.]BR ( '.R2CLI.C RD20 3KS0 S3G'~ to VAR 8340: VARIS5360' RDZDRD2 VAR i !:0. VLRIE6t0 V4RI15650 VAR 15690 ULWM'(uLLIC) ( IR'(L'C4UV) VARIE *7 DO. KS I KLS a3)K I KP a IK  NP 2 I N1PC VAR 185501 YLR 15560 VAR 1C570 VARias590 VAR 86201 VARi1Es30. ..KSPI I KL I lK z lK 1A 0.S(UftH)(UCUQ)  VL21S660.P( TJ"'. ( '(K ULH C CON'..5 )r0.ET/TADD3 VAR IS420 VLO IS440 3F~lDw.
7A 0060tMIA 0L6Sgv7A 0968LVA I 3a CA~ SimI) 0L96 t'7A O9S63tM7A j0 0 VA d061Nt) AoflsaNx0As + Od.Cw'))'A.uQ + + t (Nt)0ftl x = XIt)f ( oo.(?nOn) i13( fl2f) unc.OSZ6 M7A OdI'N Al dN O7I~t 17A OdM ( A) ONZ Nt + 1IOZN4(Vt) * + 3NZ)(11) + t bi > >4 E A c TA .(iim1 (.) .)10 6666a 01 0 MfN i NGO 07 6 "VA ZNOO NCO ZOM( + (Xt)0M £ ) O~t6tVA (~ ~ . 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 )*. (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 ±..t) 2 td9N'Zt oo0to oa 3flN tLNOD s0?_O3 M 6 dVA + NAcns + ZntO * (Nt)d(dXt)d ).% At t x NdE ONCA + lA b + A t + OZ~N(Lt) OdAN 00gla aVA 06!.(..( (nf1l)(Himlss7 (In3n)(uLM)S9V ) ZNQ.(onin) M0'a(On.o2m.onH.8~ VA 0899tMVA (oaQN 3NC647.vt Z1t0 7 Z(J~Si0(o(nsn)(a)a (1n..+fl)IdM Y(Qd.o.'N(tX) + 06Md~7A 1 Et a 6Nr NZ At 2 SW SWA r r ).
NWPC DCT 3 DIFFCO(IKL+1) DC.$ ]K .KBR DO 62100 1=2.RDX.IMKS IKt.'RDXDDCR D01(J) .KBR + K 81100 U(IK) a X(IX) C DO 82000 K=2.]K .8e1.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 .SIGt.5RDZ*( VRT * ASS(wRT) .5( U()]KS) * U(IMKS+NWPC) ) Ai(J) = 0.AeS(WRB)) . VAR19310 VAR19320 VAR19330 VLR193AO VAR19350 VAR19360 VAR1S370 VARie380 IKL a C 1 + (1)'K2NCL + (K1)NWPCL IPK .KSPI ]K = 1 + (11)K2NC + (KI)NWPC 1X a (12).ABS(VRT) ) .DIFFCO(IKL) DCL a D0FFCD(IKL+2) URT x 0.K2NC .5*RDX( ULT .5*RD2( WRT .( DCT + DCS ) C1(J) a 0.( URT EAS(URT) ) .( DCL * DCR ) C1(J) z 0.ROT * 0.IBR DO 81100 K2.SIGIArRDZD3DCT 01(J) a U(IK)ROT 81000 CONTINUE C . C 00 l8100 I=2.KER D00 82000 12.DI.X.5RD02( WRe .]K + K2NC IKP a 1MKS 1K + NWPC ]K K K2NC NWPC 1KIAS = C DCR .5( U(IK) + U(]KP) ) ULT C a 0.C1 .0..'5 WRS (W(1KMS) * W(IPKNWPC))O. = DIFFCD(IKL+3) WRT 3 (W(3K) + W(iPK))O.16PI 1K a 1 * (]i)K2NC + (K1)NWPC IX x (K2)ISR + I 1 E2100 W(]K) * x(Ix) 00 823000 K2.VRE + ABS(WRB) 1 * SIGMA*RDZD.KBR) C DO 82100 Ka2.5.5RDX( URT 1 + SIGMARDXD.1SP J a (K2)]IeR * 1 .1 ]K £ + (1])K2NC + (Ki)NWPC ) VAR 19260 VAR 19270 VAR19 280 vLRlg2O VAR 19290 VAR19300.X.ULT * ABS(ULT) 81(J) .5 AI(J) a 0.W(1K)RDT 82000 CONTINUE C CALL TDM(A .IKBR) CALL TDM(A 1.SIGMARDXD*DCL ACBS(URT) .SIGMhARDZDODCB E1(J) = RDT * 0.C1.ABS(ULT) ) .
~tSC(Hifl)ss b3(3 ).~) Al NP 1=I000ca 00 .(HL')S. ).0xCd..ZC(3aE1~S (d~t)d  00oo' ((HIiA)SS7 I )'ZQZ .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 ).v%4!)iS +.v4)tS ~m * io1.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( .fl.(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. )'ZQdS50 )Zc!S*  a ' a  (HS.Htin 0~Q0~*/'.40 C0~7A O CO0d7A NAoC1s (A(I)a . 3)3QO.mNtAz + + ONA(lt.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.L 0. z (r 01)~S~7~).LG!1CE)PA HIM.)ssv .Q ( (S'feXtiM 094=J7A OSZO0t7A (>I).'%)SS'7 + HSA .' ).3nwNtJ.
2.0.EO. IBR(K2) + I ' (1i)K2NC IK = i + (K1)NWPC IKL a 1 + (11)K2NCL + (K1)NWPCL RC a FLOAT(I1)bDX .HDX RRC a I.5RRC*RDX CFC .K2 DO 60000 13a1.IKBR) DO 84100 l=2.EO.SIE(IKP) S]EB * SIE(]KW.CF(]MKS) UC c U(IK) WC = W(!K) UL .NSEO) 33333 CONTINUE 15E0 a 0 DO 60000 K=Ki.DLeS(UL) ALC = DAES(WC) 1wS * DASS(WE) = SIE(1K) S!EC SIER .S) SIECO .X.DIFFCO(1KL*i) DCL a D!FFCO(1KL+2) DCB * D]FFCO(JKL*3) ]PK )KP a a + K2NC * NWPC = 1K .5(TS(]K)+TS(lMKS)) VAR203G0 VAR20370 VAR20380 VAR20390 VAR20400 VAR20410 .1CALI.0/RC OR = CYL0.DASS(UC) AUL . VAR20420.SIE(IMKS) SIET .NE.1SP1 00 84100 K2.5(TS(IK)+TS(IPK)) .1 (12)(KER1) 84100 W(JK) a X(lX) C 89999 CONTINUE IF(ICONV. 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 .1) GO TO 2040 ISEO = ISEO + 1 IF(ISEO.LE.Ci.0P.S]EO(]K) TSRA TSLA .NWPC CFR = CF(1PK) CFL .O.12 J .I) GO TO 55000 DCR .U(JMKS) vW = W(IKMS) AUC .DIFFCC(IKL) DCT .K2NC ]K IK 7]KS IKMS = IK .CALL TDM(A .SIE(IPK) S*EL .CF(UK) IF (CFC.D1.B.KSR IK IX a I + (I1)*K2NC + (Ki)*NWPC + K .25*RRC CYRX = CYLO0.
C*~ t )4* .I vo 0009 0:9>4 I )c t a. C0 r~t (r)LC 00~ttIdVA 060k ~wVA 0OLt dVA 00009 01 13t5.x)j! 2(AI1)3rS oo0t 9 06CI ~1VA' 09ZI dVA 0LI. 06C Zd7IA ZltrEzf 00 OOOL CO 0SCtLb7A NOI0t1D3JCZ NE d33MS CNZ 0L~tZZ7A LCECE OZ :tZ7A 1 05 3 3lNTI. (nN*31.Q ~ 000 fl co(3 S6 ..NOD (L03e.L.1! Al z(IJ)L3 (NOfl'Ol V(ISJX017) Ot)t0dVA 0001ZlA N7 dlxiol ((L1)x~JNG4(IV~~sNQ)ts0 .At~VA OSZI~7A clt'L fa ZAVtAZ'4 0=0i9 00 0=O9 00 I . t :) a .OSrlth7A OV7 t!VA t. 10 + ( 0S669 01 00 (V03VIJfl S0.+c a 21J10 to. t7) . N *:dM o zi Z)q4s)4 coo 0.± 4:7) 21J3 1u L7S1L'q'l OS 60ZMVA (Slx () I s I)SO I ((diE)l+ m1)l)'c = vzs I VIL .(r'.~t J '~1f ((t0AX'1ZJANO ) C dt0* ) * N ) S 4~ I .(N t0 *090~bVA 0O0I bVA 0O'0t MVA OSOI aVA 0OP0tZSVA (Q QtlY.OlI)03(r v+n ."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.
LE.IF C DCR OCT DCL DC8 C ( CFC.5(WCWC).5 0.i) GO TO 70000 IF(CFT.RDZ .5(DNFF2(3. 1 * 0.NU) GAMT * RPRAN .0 E2(J) .0 DI(J) D(J) + DIFFTDCT*RDZ GO 70 70000 65000 01( ) = SIE(]K) L2(j) a 0.DIFFCO(IKL+3) 1PK a 1K + K2NC IKP a !K + NWPC IMKS a 1K .NE.C2.C2(J)SIE(IKP) C2(J) = 0.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.CO(IK)0.5.1) GO TO 65000 .X GtA.0.5"(ws*Aw )'RD2 .EO.IKBR) C DO 71000 ]=m1.0 C2(J) = 0.ND.K2NC IKWS VAR21460 VAR21470 VAR21480 VA021490 VAR21500 VAR215 O VAR21520 VAR2rS30 VLR21540 VAR21550 VAR21560 = ]K ..NWPC C CFT = CF(IKP) CFB .(WC*AwCWS*iWB).1) GO TO 65500 D1(J) * 01(J) .RDZ C2(J) = 0.1.K)*DNFF2( .A2(J)SIE(3KMS) A2(J) • 0.5(TS(]K)+TS(3KP)) TSeL .(12) + K .DIFFTDCTRDZD C2(J) a 0.21730 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 .K+1))X.K2 Ix = KS.0 70000 CONTINUE C CALL TDM.0.12 DO 71000 KKI.M * * IF(TS(]K).(A2.E0.X..DIFFEDCRD2D RDT * (DIFFTDCT+D)FFS*DC6)*A02D .11.EO.TTSPLL ..0 GO TO 70000 65500 C2(j) E82(J) .I) )XMX GAMTTSTA VAR21620 VAR21630 VAR21640 VAR21650 VAR216GO0 VAR21670 VAR21680 VAR21690 VAR2I700 VAR217 tO VAR21720 VA.01.5(rS(lK)+TS(lKMS)) C GAT DIFFT DIFFB C C * RDTSIE(]K) . .1) GO TO e9950 Dl(J) = 01() .K.k)+DNFFZ(1.0.DIFFCO(IKL) = D]IFFCO(IKL+1) = DIFFCO(]KL+2) .5(DNFFZ(I.NNOPT.B2.OlFFTDCTRDZD IF(CFB EO.0 69950 IF(CFT.I DI(J) a TGA.
12 DO 33335 K=KI.09 KzK1.K) ) KIRK a 0 33329 CONTINUE C C NOTE.IK r 1 + (II)*K2NC + (K1).K2 LWPC=LWPC+NWPC AT TIMEN ) .K2 IK v 1 + (1I1)K2NC + (K1)*NWPC STOR(I.E0.NE.11.O) 00 33335 I11i. TRANSFERS VELOCITIES TO STORAGE ARRAY( 2040 KI1 K2xKBP2 LWPC=I .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.K) = SIE(IK) 333335 SIE(IK) .NWPC DO 2.NwPC 71000 SIE(]K) = X(IX) C IF(KIRK.S)=W(KK+) TO(KMS)=TO(KK+1) VAR22470 VAR22480 VAR22490 VLR22500 VAR22510 VAR 22520 VAR22530 VAR225ZO VAR22550 .5'( SIE(IK) + STOR(I.K2 1K 1 + (]i)'K2NC + (KI)*NWPC 33338 SIE(K) = 0.SIEO(IK) KIRK a i GO TO 33334 33337 CONTINUE DO 33338 1=1112 DO 33338 K=KI. 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.
'input mom'. 5. is the input to test the stability of the ADI scheme for momentum equation.2. The files.1 and Fig. The file. 5. are for calculating the numerical solution by donor cell scheme in the problem geometry.APPENDIX D TYPICAL INPUT FILES The file. 5. 5. The output results are in Fig. The output The files.6 and Fig.3(B). is for calculating the natural convection flow field in a Cartesian closed compartment. 'inputd 3' and 'inputd 7'. 5. results are in Fig.7.8. . 'inputn dat'. are for comparing Huh's And De Vahl Davis and Mallinson's correction formulas in the problem geometry. Fig. Fig. 'inputd HUH' and 'inputd DM'. 4.
0 0.0 10.0 1.0 0.0 0.0 60.0 0.0 1.0 2 5 10.rTh FILE: INPUT MOM VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 4 4 O STABILITY TEST FOR MOMENTUM 2.4 0.5 0.44 0.0 1.0 1.0 0.0 0.0 1.0 60.0 1.0 I 3 I 16G 1.0 0 0.0 0.5 6.0 0.t20 1.75 10.0 0.0 0.01 0 1.0 0.0 0.0 0.0 0.0 0.0 10.0 0.0 10.0 1.0 4 0.045 0.0 0.0 0.0 1.0 1.4 1.0 0.0 1.0 0.01 0.0 1.0 0.0 0.0 0.9 0.0 0 0 1.0 60.0 2 l.0 0.0 0.09375 0.120 1.7 0.0 2.0 1.4 0.0 0.013 0.0 4.2 10.0 0.4 5 1 1.0 0.0 0 2 1 3 O0 2 2 O O ADI 10.0 0.0 .
20 030 1.0 2.0 0.0 10.0 0.045 0.0 0.01 1.0 6 6 50.90 0.0 1.0 1.0 8 2 I 3 1 3 0 3 0 1 1.0 .0 NATURAL CONVECTION IN A CARTESIAN CLOSED COMPARTMENT 1.0 0.4 0 0 0 0 O 0 0 6 2 0.20 2 1.0 1.0 4.0 0.0 60.21795 1.0 1.0 1.0 1.0 1. 20 1.5 0.0 0 1.2 1.01 0.0 0.0 0.0 0.0 I 43 1.67 1 1I 2.FILE: INPUTN DAT A VM/SP CONVERSATIONAL MONITOR SYSTEM 'PAGE 001 5 10 0 1 0.20 1.0 32.7 1.20 1.0 1.0 1.20 0.0 0 0 2 1.44 0.0 677.0 0.0 68.+20 60.5 5 1.751.5 1.20 II I 1.0 0.0 0.
O.OGI 2.999 1.0 5.0 0.FILE: INPUTD 3 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 10 10 0 0.000 10 1.0 1.0 0.940 1.013 2.0 0.0 0.000 1.911 3 4 1.0 1. 020 2.44 4.002 0.01 0 0.0 0.0 0.0 9.0 1.07107 111 3 1 1 2 5.07107 7.0 5.511 1 1.0 0.0 1.0 0.009 G 1.0 0.0 60.0 0.0 7.0 1.0 7.+20 I 0.457 1.000 10 1.003 7 1.0 1.0 1.07107 0.0 0.0 0.0 9.70711 1.207 0.0 0.0 1.0 7.0 1.0 0.063 0.07107 0.0 2 1t 1 113 5.457 1.07107 0.0 0.0 0 0 NUMERICAL blFFUSION .47.70711 0.0 GO.0 9 10.0 10.0 1.07107 1.0 0.0 0.0 0.002 0.0 0.01 0.0 7.CASE 3 2.0 9.014 ).045 1.0 1.75 0.0 0.0 0.9 0.497 2.0 1.05.02 1.0 9.991 6 1.0 0.07107 12 12 Ill 2 5.5 1.409 1.3 1 I 1.0 0.009 4 1.5 0.0 1.0 0.751 2 1.948 1.0 0.0 4 I 36 1.207 0.0 0.0 1.II I 11 2 2 1.000 9 1.5 1.0 0.0 9.0 2 1.O 2 1.0 0.0 12 11 2 121 t 5.0 1.000 0.061 1 0.0 7.0 2 4 2 4 1.0 1.249 3 1.0 .0 5.204 0.09375 0.0 0.0 0.0 0.000 0 0 0 O .0 1.0 1.001 0 1.063 ( ).997 7 1.0 0.0 0.0 0.026 5 1.0 0 2 5 3 0 7 6 1.0 0.0 8 10.000024 60.0 9.000 0 0 O O I.0 0.974 5 1.204 0.0 0.0 0.497 0.000024 0.000 0.0 1.0 1 .014 0.
003 17051 1.0 0 0 2 2 1 0.07 07 0.+20 2.07107 I 113 7 2 1.0 0.003 0.0 1.0 3 0.0 0.0 1.0 2 9.0 2 7 8 5.0 1.990 5 1.457 2.0 0.000 6 1.045 1.5 1.0 0.00.000 0 O 0 0 0 7 1 2 7 2 5.0 NUMER ICAL DIFFUSION .0 0.0 I 1.0 1.014 0.0 4 2 4 1.07107 7.75 2.001 0.0 1.0 10.333327 60.9 0.0 0.01 2.0 10.07107 1 1 713 2 5.0 5.143 0.0 0.0 0.3 1 1 1.0 0.44 4.17051 0.000 0 0 0 1.CASE 7 0.420 1.143 0.0 0.00.911 9.0 60.0 1.000 6 1.013 2.0 1.333327 1.47.0.FILE: INPUTO 7 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 6 0.09375 .5 1.0 60.633 1.0 0.0 7.0 1.0 0.0 7.014 0.374 1.0 0.01 0 0.0 0.0 1 9.0 7.0 0.071107 0.0 1.0 5.0 0.0 1.4 8 2 1 3 0 0.0 0.07107 0.0 4 I 36 1.0 0.009 1.0 1.0 0.0 7.0 0.0 0.457 2.0 4 1.0 0.0 0.0 0.0 7.0 9.0 1.013 4 1.001 0.4 1.0 1.07107 711 8 8 2 5.0 1.0 0.167 1.0 0.0 3.01 0.0 0.0 2 0.0 1.0 .0 0.002 5 1.626 1.0 0.0 1.633 1.907 3 1.0 2 2 0.0 5.0 0.0 0.0 1.
0 01l 0'00' C0'I C.0 CGO.NOIS(fJJIG IV:)IUIW(IN too 39vd IVNOIIVSU3ANIX) dS/W~A v VIIi III)II OIMINI O~N :311A II .09 0.0 0 *I O'l 0' *1II1 01l 0*1 0 1 0'L 01 O'1 0..u *0 Ul' I 6P I 0011* I 0. 1'0 0.1 1' 0*0 0 *1 D99C00 0'1 0IO' 0*0 0*0 0*0 0'0 0 *I0*0 9c 1 t.09 W '09 0.0 GLO10 ZU10 SL9*0 109.0 09G"0 9Cs£0 661''0 6sv1'.SAS 3~~~)Vd kOIINOW 0 0'0 0*0 0.*I riI 00 0*0 0*0 0i'00 0 10*0 6 v 0*0 0'0 top' .0 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 .L Z01 'L L'09C91C P09c91''? 1'Q9C9p 0'19C91' 1 iLO'O 1'LO*0 GL00O LLO'0 £00. *0 PI 9.z 10'o 0 1 0'l GL'O 1O1 '0 I 0*0 01l 0'l L910* 199*0 099. 0 LG0 . 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 9?LL '0 G'.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 ~W31.1bl b 690*6 £00*0 LC9 0 90c.
0 8.20 911 1.20 10 1.20 5.0 26 26 5.20 1011 1.6603 0.0 2 25 25 5.0 0.0 I 1 5.20 2 1.0 5.20 I13 1.20 1 1.0 0 2 1.0 .6603 0.0 0.FILE: INPUTO IIU4U VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 2 0.0 0.20 2 1.0 2 25 5.6603 0.20 9 1 1.0 5.20 913 1.
0 0.013 2.302 0.0 1.0 1..093 0.0 0.002 0.120 2.920 0.566 0.0267 0.522 0.0 0.030 0.0 1.007 0.0 1.09375 0.0 0.0 0.099 0.0 0.0 1.072 ' II 12 13 14 15 16 17 10 19 20 21 22 23 24 0 0 0 I '.0036G0 0.767 0.0 0.0 1.463704 9.02 2.02.2406 1.0 0.949 0.496 0.4 4 1 36 1..0 1.0 1.0 1.045 1.034 0.0 1.242 7 1.0 60.33G 6 1.0 0.700 .0 1. 1.0 0.002 0.0 I 1.460 4 1.903 0.3 1..OM 0.636 0.0 0.0 0.0 0 0 NUMEtICAL IrFFUSION .96G 0.0 0.0 0.0 1.365 9.144 1.223 3 1.0 0.0 5.030 0.197 0.0 0.567 9.051 0.20 1.0 1.0 0.016 0.0 0.003 0.44 4.5 1.0 1.0 0.0 0.029 0.0 0.0 1.0 0.0 1.246 0.0 0.0 1.951 0.705 0.0 0.4 0.174 1.011" 0.932 1.0 1.0 1.5 1.420 0.0 0.30i9 1 0.02.463604 2.0411 0.260 0.4GI6 0.0 0.0 1.009 0.026 0.0 1.115 0.0 1.01 0.0 0 2 1 3 0 7 6 I 0.GOG 2 1.75 1.0 0.044 9.360 0.005 0.463604 2.0 1.OGI 0.0 2 2 2 0.526 3 4 5 G 2.0 1.5 1.0 0.0 2 4 2 4 1.9 0.003 0.156 0.0 5.0 0.463G04 2.0 1.0 5 1.01 0 0.000 0.0 0.003660 60.0 1.133 10.0 1.0 1.0 3 10 1.969 0.403604 2. 4G1IG04 7.021 0.002 2.076 0.0.003 0.0 60.0 0.rILE: INPUIO OM A VM/SP CONIVEIlSATIONAL MONI(fro SYSTEM PAGE 001 24 f( 0 I 0.0 0.
SILr.20 1011 1.0 0.20 I 2 1.20 913 1.20 1.0 I*  .0 0 29 25 I 25 2 1.6603 0. GG03 0. 6603 0.0 2 5.0 5.20 911 I.0 0.0 26 2 1.10 OM A VM/SP CONVERSATIONAL MONITOnr SYSTEM PAGE 002 2 0.0 26 5.0 5.20 113 1.0 2 5.20 5.0 0. & .: INP.0 I 5.20 9 1 1.0 .