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
2.3 Low Reynolds Number Flow CHAPTER 4: NUMERICAL DIFFUSION Truncation Error Diffusion 4.2 4.3.4.3. Implicit and 4.3 Basic Assumptions and Limitations 2.2.1 Mesh Point Implementation Mesh Interface Implementation 4.1.4.3.3.3.2 Truncation Error Diffusion 4.2 ke Model 3.1 Truncation Error Diffusion in a One Dimensional Problem Another Approach for a One Dimensional 4.3 Discussion of Skew Differencing and Corrective Schemes Raithby's Scheme 4.3.2 Energy Equation 2.4.5 SMAC Scheme and Compressibility CHAPTER 3: DIFFUSION MODELLING 3.1 4.3 Tensor Viscosity Method Corrective Scheme by Huh 4.2.3 Truncation Error Diffusion in a Two Dimensional Problem 4.4 ADI Schemes Conservation 4.1 Assumptions 2.1 4.3.2 Crossflow Diffusion 4.1 Continuity and Mass Diffusion Equations 2.2.2 .4.2 Physical Constraint 4.2 Limitations 2.1 Skew Differencing by Huh 4.1 Diffusion Convection 4.1.2 Comparison of Explicit.2 Governing Equations 2.2 Turbuletit Diffusion 3.2.4TABLE 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 2.1 Derivation of Turbulence Equations 3.4 Solution Scheme 2.2.1 Problem Statement 2.4.1 Molecular Diffusion 3.4 4.4.3.1.
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
. . skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with 6 = 63.3(B) .12 . .11 Comparison of the true solution and solutions by donor cell scheme.. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure . numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig. by De Vahl Davis and Mallinson's formula along the line BD in Fig. . .3(B) . . . . ... 137 5. .3(B) . 5.43* . . CA in Fig.10 Comparison of the numerical solution with donor cell differencing of convection term and analytic solution with increased diffusion constant. 142 5. 144 . . . .13 139 5. 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... . . .. .convection problem with 6 = 450 Comparison of the true solution and solutions by donor cell scheme.15 . Donor cell solutions with various diffusion corrections for a pure convection problem with 8 = 450 ..9 5. problem with 6 = 63. 135 5. .. True solution and donor cell solutions with and without diffusion correction for a pure convection . . . . .14 140 5. (D+DDM).16 143 5. 5. . Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 6x6 meshes along the line . .17 Comparison of the analytic solution. . . . . . 5.. . 5. 136 5.430 ..8 . .
. 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. 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.. . .... . . 5.. 152 5.... . Simple recirculating flow field for test of the implementation strategies of the corrective scheme . .23 . . .. . .10 5. . . . .18 Comparison of the analytic solution. . . ... . 148 5. . .. ...19 .. . . . . .. 146 5. Comparison of the true solution...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 .. .. . .3(B) . .. .. ... . 5.19 ..20 .. .147 5.. ..22 150 5.. numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig. 145 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 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 .. 151 5. 153 . .24 .21 S.19 . .
. .3 sec.5 sec and the Courant limit of 1. . .3 sec. .. with donor cell differencing of convection term without any diffusion correction . corrective scheme .31 1i60 5. . 0. 156 5.. . 5. . . . 155 5.27 .. . with mesh interface implementation of the . .32 161 5..cx T where 4 is given by. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. . 163 A. 1 178 . .33 Comparison of the ADI solutions for the profile of 4 at t = 10. ...0 sec along the line AB in Fig. . .1I 5. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. 5. . 157 159 5. 5. . . .. .0 sec ..0 sec and the Courant limit of 1.30 with . 5. . . . .0 sec . Problem geometry for comparison of the explicit and ADI schemes .30 5. different time step sizes Boundary conditions in terms of 4 for the problem in Fig. . . .3(B) = e. ..30 for At = 1. . Steam concentration distribution after one time step..30 for At = 0. . . . 0. Steam concentration distribution after one time step.28 . . .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 .. . .. .3 sec. with mesh point implementation of the corrective scheme ... . . . . . . . .29 Steam concentration distribution after one time step. . . 154 5.. 0.
. . . . .114 . . . Page .12 LIST OF TABLES Number 4. . . . . . Comparison of the Von Neumann analysis and numerical experiments for the stability condition in terms of the cell Reynolds number in an explicit. . . . . donor cell differencing of the convection term . 78 4. . 1981 . . . . . . . . .2 .1 List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research. 2D. . .
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.C x y D D ah D as Dhs D .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./Imaginary (1) Turbulent kinetic energy (2) Thermal conductivity (3) Ratio of specific heats. 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 .d xy e f Gr h I Im k g M Nu P Pe p y Az Pr Q Prandtl number Heat source .D d .
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.1 v u Mesh configuration. tan.
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 .
From the hydrogen transport point of view. The governing conservation equations are decoupled in order to simplify the solution procedure. The error due to decoupling is negligible in a slow transient where the . Some simplifying assumptions are made concerning the thermodynamic state in the slow mixing stage. air. hydrogen.I~~U1i* 11*11044~1 1~. The four components. 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. although it may be less than 100% when there is no liquid component present. The distinct feature of the second stage is a much longer time scale in comparison with the first blowdown stage. 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. it is necessary to understand such problems. the fast blowdown stage and the slow mixing stage. Some remedies have been proposed to deal with However.( 1~111111_~. the fluid dynamic phenomena in the containment in order to justify those remedies. the response of the containment during an accident can be divided into two stages.
Finally. and the total diffusion constant is the sum of the diffusion constants of those two mechanisms. Diffusion occurs by two independent mechanisms. of each component in multicomponent gas depends on the mole fraction of that component. the phase change is taken into consideration to maintain 100% relative humidity. c. In the first step the continuity and momentum equations are decoupled from the energy and other scalar transport equations and solved by the Simplified Marker and Cell (SMAC) method in order to obtain the flow field. Convection and diffusion are the central issues in physical modelling efforts of hydrogen transport. resulting in the same convection velocity for the four components. The molecular diffusion constant is predicted by Wilke's formula [73) for multicomponent diffusion and ChapmanEnskog formula [5) for binary diffusion. The diffusion constant. and turbulent dissipation rate. Convection is assumed to occur as a homogeneous mixture. The turbulent diffusion constant is predicted by the kc model [45). The turbulent kinematic viscosity. molecular and turbulent. The turbulent kinetic energy. .17 state change over one time step is small. k. 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. are determined by their own transport equations.
The numerical diffusion has two different sources. The turbulent Prandtl and Schmidt numbers are assumed to be equal to one. The effective diffusion constants of the two errors are of the same order of magnitude. 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. truncation error diffusion and crossflow diffusion. 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. Truncation error diffusion occurs in the flow direction while crossflow diffusion occurs primarily in the direction normal to the flow. Since the error usually appears as an additional diffusion. The leading issue in numerical modelling of convection and diffusion is to minimize the error that occurs in the numerical solution procedures. however the latter turns out to be the dominant error source in most convection dominant problems. numerical diffusion. Therefore. has been used to describe the numerical error in general. There has been much debate on the numerical diffusion and many schemes have been suggested for the past two . the term. Truncation error diffusion is a one dimensional profile error and crossflow diffusion is a multidimensional operator error [68) of the finite difference equation.18 which is the diffusion constant for momentum transport can be calculated directly from k and E.
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. 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. The corrective scheme is inherently better than the skew differencing scheme in that it is conservative and does not affect the simple solution procedure. skew differencing and corrective schemes. However. ADI and implicit schemes. 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. Raithby's [53) and S. The corrective scheme can be implemented with any of the explicit. The schemes presently being used can be divided into two categories. The conservative property is essential in a long transient problem like the hydrogen transport in the containment.19 decades. have been tested for recirculating flow 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. Two implementation strategies for the corrective scheme. although with different stability conditions.
[49) includes all the Development Laboratory) [6]. simulation results compared with experimental data.20 implementation has always given physically reasonable solutions and may be used extensively for all diffusionconvection problems. The suggested physical models and numerical schemes have been used to simulate the LOCA experiments performed in BattelleFrankfurt [44) and HEDL (Hanford Engineering Ref. .
e. a large amount of steam and water will come into the containment in the fast blowdown stage increasing the containment pressure. In addition to the pressure increase due to the primary coolant.1 Problem Statement Once a lossofcoolant accident (LOCA) occurs in a Light Water Reactor (LWR). In order to justify the design of any mitigation system. hydrogen generation gave a serious concern about the integrity of the containment in the TMI2 accident. it is essential to understand the fluid dynamic phenomena in the containment. containment inerting.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. There have been some mitigation procedures suggested. . Thereafter.g. for dealing with this problem. installation of ignition devices.. use of flame suppressants and enhanced venting capability. The major safety concern is that of keeping the containment pressure below a certain level to prevent a large scale leakage of radioactive materials. the hydrogen has received much attention in the safety analysis of nuclear power plants.. After a while the slow mixing stage follows the initial blowdown stage and continues for an extended period of time.
. some simplifying assumptions should be made to use the numerical procedure without impairing the acceptable solution accuracy. laminar and turbulent flows.g. etc. . The most difficult aspects of this analysis are the complicated geometry and chaotic postLOCA conditions. thermal nonequilibrium._ __ _ __ _ I_ ~ 22 Since a numerical method is suitable for this purpose. e. phase change and heat transfer between the gas components and wall. a good computational tool has been required to predict the hydrogen concentration distribution. Therefore.
(2.6) Equation of state: p = f p (ii .4) Turbulence kinetic energy: Dk 1 t 3k 3 P 1 k ok x .5) Turbulence dissipation rate: D 1 _ 1t StE xk Dt P Txk 0E ] + C t*( 1 p k ui xk ouk au.3) Mas s diffusion: S+ V*vp i . The physical implications of the conservation equations will be given with their basic assumptions and limitations.1) Momentum: pgI + d + V0 (2.S + + t Bui P au k aui + ax~ ) x k X axk gk Pr t t 'T xk (2.2) Energy: p[e e + Vvh] + = VkVT + e fg (2.2 Governing Equations The governing conservation equations are set up to describe the postLOCA fluid dynamics in the containment.t = VD. Continuity: V'v= 0 pl + Vvv ] = Vp + Vv = 0 (2.Vp i + 4.7) .xk k 2 v t BT C2 + kg k Pr 3x 2k k t (2.23 2. I+ 3x 1 .T) (2.
= D. Im + Dit it (2. The subscript i denotes the four components as .8) f (v) (2.T) (2.17) t = C k 2/C E (2.t (2.13) D. .16) kt = .12) (2.T) (2.11) k = .10) v = vm + Vt (2.24 Constitutive equations: 'N = f (p .k + k t m Dim = fD(Pi'T) (2.18) p = Epi (2.15) Dt = v t (2.14) m = fk (PiT) (2. = f (Pi.9) D.19) Note that 1.
w. there is one moreequation than is required.k C for a three dimensional case.20) i = 2 = 0 43 + 04 = 0 (2. The phase change occurs between steam and liquid droplets and the diffusion constant of liquid droplets is equal to zero.i (4). v. p.2. one continuity. three momentum. air. 2.1 Continuity and Mass Diffusion Equations The continuity equation is redundant with the four mass diffusion equations and there is'an inconsistency between them.25 follows: i=l i=2 i=3 i=4 2. Therefore. four mass diffusion and two turbulence equations with ten primary unknowns. The mass diffusion equations are summed up for the four components. The turbulence equations are written in terms of the Cartesian components in tensor notation. u. i=14. 3. .22) There are eleven conservation equations. one energy. liquid droplet. (2. H2' steam. T.21) D4 = 0 (2.
Therefore. 2. Since the concern is only for the slow mixing stage after a LOCA. at p= 0 (2.(Ep i ) a .26 .23) The sum of the phase change terms is equal to zero.25) Vp i = 0 (2.27) Consequently Eq.1 and Eq. 2. Vp = 0 (2. it is reasonable to assume that Eq. ap at S+ Vp = E(VDiVp i ) 1 (2.24) Now it can be seen that the continuity equation is valid if and only if the following Eq. 2.) i + E 1 (2.27 remain valid throughout the transient. Eq.26) It is a reasonable assumption in a turbulent flow regime that the four components have the same diffusion constant. .4 are consistent only if the temporal and spatial variations of the total density are very small.) + V(Ep Jt i = (V'D. 2.Vp. 2. 2.26 are satisfied.27. 2.25 and Eq. 2.25 and Eq.26 can be reduced to the following Eq.
(p. [c pihi .2 Energy Equation The energy conservation equation. p + £e.30) Eq. Be. 2.3. I at t + V [vp.31 so that the internal energy and enthalpy can be written in terms of the temperature.29) The second term on the left hand side of Eq.i = V'kVT (2. It is clear that the major contribution to this term will come from the latent heat of the phase change.27 2.ih. E it p + e. Eq. 2.30 is substituted in Eq.iei) + V'[Ep.i + V. LeaPi LeQ.29 denotes the energy change due to the concentration change in a given control volume.2. = VkVT (2. 2.29 to give the following.28) The temporal term can be divided as follows. is also of an approximate nature and the exact form is given in the following. ae. 2.31) The thermal equilibrium assumption is incorporated in Eq.2. .ih i ] = V*kVT (2. (2.
Pic vi T c p p.T P P e Piei p ih p.34) 2.c .33) h (2.28 (Picvi )~ +V"[vTZpic pi] + efg 9kVT = (2.T 1 Eq.T h. P (2 33) (2.1 = c Vl.3 Basic Assumptions and Limitations The governing equations are based on some simplifying assumptions about the physical phenomena in the containment after a LOCA. = c p1 . 2. It is necessary to clarify these assumptions and their limitations for a safe use of the given physical models and solution scheme. 2.32 become identical by defining average internal energy and enthalpy as follows. .32) where ei i = eg4g + e 4) Sefg g e.3 and Eq.
vt. 3. The change of the diffusion constants. 5.1 Assumptions 1. .3. The phase change rate should be moderately small because the solution scheme decouples the phase change term from the energy conservation equation.2 Limitations 1. with no slip. 2. 2. The four components all have the same convection The liquid droplets move with the gas components velocity. Dt and k t over one time step should be small because the turbulence equations are decoupled from the velocity field calculations. The relative humidity may be less than 100% when there is no liquid component left. The relative humidity in the containment is 100%. 3. The turbulent Prandtl and Schmidt numbers are equal to one.3.29 2. 2. 4. The four components are treated as being nearly incompressible and the phase change rate is moderately small. The four components are in a state of thermodynamic equilibrium. The temporal and spatial variations of the total density should be negligibly small because of the incompressibility assumption.
i Since it is too difficult to solve the whole system of the equations and obtain a consistent solution for all the unknowns. k. Eq.5. 2.1Eq. In the first step the SMAC scheme is used to get the convergent velocity field with zero divergence for every mesh. the conservation equations are decoupled into two sets. The equation of state. continuity/momentum equations set and other scalar transport equations set. energy equation and two turbulence equations.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. T. 2. .4 Solution Scheme Eq. It solves the . E. Pi(4). The phase change is assumed to be zero in this step. the relative humidity may be less than 100%.7. p. The SMAC scheme is In the second step. v. the obtained velocity field is substituted in the scalar transport equations. . explained in detail in section 2. In the third step phase change is taken into consideration to maintain 100% relative humidity. 2.30 2. 2.7 are a coupled set of equations with primary unknowns. which are the four mass diffusion equations. is used to update the reference pressure and reference density which is important in calculating the buoyancy force. When there is no liquid component left. u.
ay (Eq. Since the continuity equation used in the SMAC is an incompressible form.36 and Eq. it is possible to modify the SMAC scheme to accommodate a slight compressibility effect. 2.36) ydirection momentum: av av ua av 1 + pg + a y x2 Nv 2 (2. However. 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. which can arise with net inflow or outflow boundary conditions. i..ax 2. Continuity: au ax av ay 0 (2.e.37): . the SMAC scheme can be used only for incompressible flow calculations.37 as follows.37)y (2. (Eq.37) Py The vorticity conservation equation is derived from Eq. the divergence of the velocity field vanishes everywhere.36) . 2. The continuity and momentum equations are given in two dimensional Cartesian coordinates in the following.1_ _~ 31 continuity and momentum equations to get the velocity and pressure field in an iterative way.
36 and Eq.40) S (2.41) where fn and gn are the explicit quantities at time step n. 2.39) where av au D= . 2.37 as follows.36 and Eq. 2.32 a C 2 a2 + u+ V = v+ ) at . 2. (Eq. In the tilde phase the approximate velocity field at time step (n+l) is obtained from Eq. ax 2 2.36) + ~(Eq. ay 2. the socalled tilde phase and pressure iteration. .+ y ax ay The basic idea of the SMAC scheme is to separate each calculational cycle into two parts.37): a2 = 2 ax 22 axay a 2 ay 2 (a2D + a2D aD +1 2 at Re ax 2 ay (2. 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 (2.38) =ay ax The Poisson equation for the pressure field is also derived from Eq.37 as follows.x ay ax 2 ay2 where au av (2.
2.41 are still unknown. 2. 2. V2p = f(u. 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.44) Eq. 2.44 is given in the following. a guessed pressure field or the pressure field at time step n is used to start the iteration. au ay au av ax  (2. _ Pi+j2P +P iPij+ pij f(uv)  Dn + l D n D D Ax 2y t (2. The equation of vorticity transport.38. follows. the following relations hold after the tilde phase. Therefore.42) av .v) D (2. Eq. 2. 2.43) ax ay It can be seen from Eq.39 that the pressure field is not right due to the fact that the divergence of the velocity field is not equal to zero.33 Since the pressure field in Eq.40 and Eq.40 and Eq.O 0 (2.45) .41 have the right vorticity. 2.39 may be rewritten as A finite difference form of Eq.
. 2.36 and Eq.34 The pressure correction formula is obtained by conn+l only and putting D.48) The SMAC scheme can be extended to a slightly compressible fluid flow. The slight compressibility means that the NavierStokes equations may be solved safely with the incompressibility assumption. (6u) L =  pAx (6u) R = t6p pAx (2. S+ at V"(p) = 0 (2.37 so that it does not change the vorticity implemented in the tilde phase.47) ) The pressure correction is given in terms of the residual divergence for every mesh. equal to zero. no more correction will be made.49) . sidering the cell (i.46) (2.j) 26p( 1 Abx 6p = h + _) Ay 1 At( A + 1 Ay 1 Dt D (2. The velocity correction formula is derived from Eq. while the continuity equation has the following compressible form. Once the zero divergence velocity field is obtained. 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. 1 . 2. Now the problem is how to converge to a predetermined nonzero divergence field.51) v p where m.52) (2.50 may be given from the previous time information and boundary conditions. The associated physical assumption in Eq. : in n =Vp v  total net inflow of mass at the boundary during the time At.49 the nonzero divergence may be derived as follows. 2. D = V'v 1 p p at V p (2. V: total volume of the containment.35 From Eq. .p_ p at 1 min/V At n p n n (2.uniform over the whole containment. 2.50) The two terms on the right hand side of Eq. This can be done with the following modification of the pressure correction formula of the SMAC scheme.51 and Eq.
The velocity correction formula remains the same as Eq.(DD 1 1 + ) 2At( Ax Ay (2. 2.53) where D O is given by Eq.6p = .48.50. 2. .
air. 3. Although turbulent diffusion is greater than molecular diffusion by a few orders of magnitude.37 CHAPTER 3 DIFFUSION MODELLING The hydrogen. components. As collisions occur more frequently. the process Since there may be three gas of diffusion is also increased. in the containment after a LOCA. hydrogen. The thermal conductivity and viscosity are also diffusion constants in a broader sense for momentum and energy transport. 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. which should be modelled independently. Convection occurs as a homogeneous mixture resulting in the same convection velocity for the four components.1. It gives the diffusion constant of each component in terms of . molecular and turbulent. the diffusion constant of each component is calculated separately by Wi~k's formula [73] in Eq. Diffusion occurs by two different mechanisms. air and steam. steam and liquid components are transported by convection and diffusion in the containment. 3.1 Molecular Diffusion Molecular diffusion occurs by the collisions of gas molecules.
38 the binary mixture diffusion constants and mole fraction of that component. Eq.1) where D.2. T3 D = 0.. 3. stants are given as follows.: yi: Diffusion constant of the ith component Mole fraction of the ith component D.2) poAB D.AB where DAB: AB Binary mixture diffusion constant in [cm /sec] .: Binary mixture diffusion constant between the ith and jth components. The binary mixture diffusion constants are calculated by the ChapmanEnskog formula.0018583 1 + M A MA MB (3. 1Y A B DAB 1Y B yA DBA 1Y C YA CA C YC CB B YC DBC C D = C YB DCB (3. The property of air will be replaced The resulting diffusion con with that of nitrogen gas.
3.lll~*. (3. 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.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.AB: Dimensionless function of the temperature and intermolecular potential field for one molecule of A and one of B. =n [ p n = [+ S n . the following relations are required.5) 2 n . P n ]. OAB = (o A + oB (3.3) CAB = AB For the pair of polar and nonpolar gases a correction factor is introduced to account for the polarity..3.4) np = np 2 where the correction factor 4 is given by. In order to calculate the para meters aAB and 0DAB in Eq. ap = np £ (o n + o p) EE e (3.2. 3.
6810 x 10 .= 1. H2 a n = 7.2 E/K = 380 0 K a = 2.9492 x 10 D ah D 1 T .6 x 10 The fitting functions for the binary mixture diffusion constants are obtained by the leastsquarefit method.65 A The polarizability is given for the hydrogen and nitrogen (air) components. 3.6 T1.6947 as .5 are defined as follows. t* .40 The parameters in Eq. The containment pressure is assumed to be1 atm. 5 = 4.9035 = 4.9 x 10 25 25 25 [cm3 3 [cm ] : N2 : an = 17. 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.
i (3.6) The viscosity and thermal conductivity can be considered as diffusion constants for momentum and energy transport and calculated similarly as follows [5).__Li~n_~PIIII ~.7) k.lllI_ 41 Dhs = 2.8144 (3. = n i=l y.10) Viscosity in [poise) Temperature in [K]) LennardJones parameter in [A) Dimensionless nTumber . n mix = Yipi n i E j=l Yj ij (3.67 x 10 where : T: a: Svi : o2 2 (3.k.lii. = 2.8) j=l where ij 8 Mj j ij M The viscosity of nonpolar gases may be obtained by the following [5).8916 x 10 where T: D: [OK1 [cm 2 /sec] 5 T 1.
385 nn y. Wsteam = 0.[5].42 The viscosity of steam is. 3.12) where c P is the specific heat per mole at constant pressure. The ks model sets up the transport equations with some empirical constants for the turbulent kinetic energy. seems to be the kc model originally developed by Launder and Spalding [45).2 Turbulent Diffusion The diffusion process is greatly enhanced by the turbulence of fluid flow. E.y RT k=l PMiDik k i There is a useful relationship nonpolar polyatomic gases.11) for the Prandtl number of Pr = c (3. The . k. 200 C There is another formula suggested for the viscosity of multicomponent gas in the following [5]. c (3. The effective diffusion constant is therefore the sum of the molecular and turbulent diffusion constants. There are several models suggested for Presently the best model calculating turbulence effects. and turbulent dissipation rate. 2 n Yi umix = C 2 i=l Yi + 1.009 cp at 1 atm.
etc.= v ( Dt + ui j u P= Dt ax x 2 k6 ij 3x_) 3 i (3. u. = u.' etc. + Pi Pi = Pi + p. 3. + u.43 turbulent viscosity is given directly in terms of k and C.15) .u. momentum and energy equations by decomposing the variables into mean and fluctuating parts and averaging the resulting equations.u. The effect of turbulence appears as the additional terms due to the product of fluctuations which do not necessarily cancel out. . 45] The transport equations for the turbulent kinetic energy and turbulent dissipation rate are derived from the continuity. 29. u. 34. (3.' j T'= a t x.' For example. These terms are treated by the eddy viscosity concept of Boussinesq and eddy diffusivity concept which are given in the following.1 Derivation of Turbulence Equations [15.2.14) .13) (3.
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. 3.2.16) .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 ] .2 ke Model The transport equations for the turbulent kinetic energy and turbulent dissipation rate form the basis of the kE model. j=i. A two dimensional cylindrical coordinate form of the kE model is introduced in Eq.44 The transport equation for u. 3.' u is transformed to the turbulent kinetic energy equation by the contraction.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.
3.metry.09 C1 1.44 C2 1. Since the kE model is applicable to fully . 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. 3.17.45 The turbulent viscosity pt is given in terms of k and E in Eq. = C pk 2/ (3.0 o 1.92 ok 1.3 The molecular effects can be included in Eq. 3.16 as follows. C 0. it is difficult to determine whether a region under consideration is in a turbulent or laminar flow.17) The suggested values of the constants are given in the following table.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.
18) In the above equations C 1 1 ok and a retain the values assigned in the ordinary kc model. D Dt p ax k [(t Dt+ ) e k + C k i + k x ) au axk C 2 E2 2 t 2.2vak 2() k 2 (3. 41).5/(l+Rt/50)] C2 =2 where R 2 0 [1. Eq. Rt. 3.0 . CP = CP0 exp[2. A more refined model is introduced in this section because consistent treatment of laminar and turbulent flows and transition between them may be required in the future modelling efforts.lit k + l ak aui i + t (ax k + xk auk axi axk e .3 exp(R2t ) (3. it is required to extend the model to laminar flow regime or to switch to laminar models according to some criterion. The model presently being used is just to use the sum of the molecular and turbulent contributions for every region in the containment.19) = pk2/VE .16.0 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. while C1 and C2 are to vary with turbulence Reynolds number.0.46 turbulent flow.
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)
7. It can be extended to a two dimensional form as follows.y+b) = f(x.y) + (f) af (4. Eq. 4.y+b].8) 3 @n + u()n + a( ax i+b ax 4 n+g _ n )n+g .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. n+g 4 i+ Ax n i Ax 1 xji+ n i+l .51 Subtracing Eq. we obtain the error at time step (n+l).a(?) ) 2(. . 4. n+l i+f EXACT n+l i FD n i+f  4P n a n+g i + At [ u( x ) i+a (4.i+c T i+d ax where the following relations have already been used. af f(x+a. 1 < c < n 2 4 + i Ax 2 11 4 2 1 i+d < d < 1 The mean value theorem.x+a) and [y. is for a one dimensional case. 4.3. n .6 from Eq.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.
F (4.10) The derivatives with respect to time and space in Eq. 4. 4.8 to the following form. . Although the error terms in Eq.xi+ ] and [tn tn+). E . they may represent the errors over many time steps in a stable scheme.C ax 22 +At g t . 4.tax . D and F are time derivative terms which become negligible for slow transients. 4.D +aAx(cd)Atax 2 +agt .A fAxAt axat 2 B utx(a+b) .9 reduces Eq. C and E are truncation error diffusion terms while the terms B. 2 n+l (4i ) EXACT  (4+) FD = i n+1 f )2 2fx 1f 2 ) (f 2 2 . In a stable scheme the error introduced at a certain step decreases in its absolute magnitude in the following steps and most of the error occurs between neighboring time steps.10 are at some appropriate points in the domain [xi . The terms A.52 Now Eq.10 are based on the assumption of perfect information at time step n.
The analytical solution of Eq. 4. The results may be extended to a general one dimensional problem if the effects of the transient and source terms are not dominant in determining the profile of 4.12) and C 1 =N cL 1 e C2 ecL 1 . 4= Cecx + C2 where c = u/a .53 4.11 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. cL (4. u ax a22 (4.1 Truncation Error Diffusion in a One Dimensional Problem The truncation error diffusion in a one dimensional problem is analyzed for a steadystate case with no source.11) ax Consider a case in which the domain is divided into N equal meshes and the mesh size is Ax.
i+l i = (P+l)( 1 . i+  i = (P+l)(4)  4)i (4. 4.13 can also be solved with the given boundary conditions by reducing it to the following form.16 is easily obtained from Eq. i u x i+l2.11 is finitely differenced using donor cell scheme for the convection term in the following. Equation 4. 4.15 and solved for in subsequent procedures.15. P(4). Equation 4.15) Equation 4. 4.16) .54 Equation 4. where P =u a  l i24 + ii (4.40) . (4.14 may be recast to Eq.i + 4i2 Ax = a (4.14) The parameter P in Eq.13) where the velocity u is positive.14 is the cell Reynolds number or cell Peclet number according to whether the diffusion constant a is the kinematic viscosity or thermal diffusivity. .
4. (1+P) 1 .20) Now the analytical solution Eq. 4.1 Inserting Eq.'N1 (1+P) N1  ) 0 N . 4.12 and numerical solution Eq. i  0 = 4.LI II .I_.~LI LIII~.20 have been obtained without any approximation for Eq.1 (1+P) N 1 (4.16.17.11 and Eq. (1 0 ) is ) P( N ¢0 (1+P) given as follows.17) 01 can be expressed in terms of 40 and 4N N N by Eq. as. 4. 4. Both the analytical and numerical . 4.19) (1 . 1 (1+P) i . 2 1 = (1+P) (4 0) 4' N 92 = N2 (1+P) 2 (  0) = N.(1+P) P (1+P) 0 + N . 4. we obtain the solution for 4.41 = N ) (1+P) [(1+P) N1  2) (4.13.19 into Eq.~l*IIIIX1~~^ ~ ^I ^~Xi 55 Then.18) (l+P) + 1 Therefore.
21 and Eq.22 we can obtain the effective Peclet number Pe of the finite difference solution in terms of the real Peclet number P as follows. 4. Eq.13.22) Comparing Eq. 4.23) (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. in Eq.25) .+1 i + ) il 0 (4. The in analytical solution Eq.13. 4. and the effective diffusion constant for the numerical solution will be obtained. 4.1 e Therefore. i+l 2Ax il S S2ax 4) i+l 24i + i 2 i (4.24) It is also possible to go through the similar procedures with a central differencing form of the convection term in Eq. 4.20 is also substituted for 4. 4.1 (4.25. 4. e (il)Pe (ePe1)2 NPe .13. 2 ii P ( (1+P) (1+) N _ 1 N 4).25) (4. Pe = in (1+P) (l+P) iIp2 N (1+P) .56 solutions will be reinserted to the finite difference equation.12 is substituted for 4i Eq. 24 j. 4.
4. it helps understanding the origin of the truncation error diffusion.24 and Eq. 4. Therefore. The convection terms in .26) where 2 < P < 2 The effective diffusion constant De can be readily obtained from the following relation.25 can be derived as follows. Pe = kn (2P) 2+P (4. uLx D De Pe  (4.57 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. 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 4.1.27) In general.26 may be a good indication of the truncation error diffusion occurring in a general one dimensional problem. Eq. 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. Although this approach is approximate and overpredicts the truncation error diffusion.
4.29 hold only when the Peclet number is large. 4. 4. 4.( i+f 2 (4.29) i+ The truncation error diffusion can be quantified by subtracting Eq. ae u( x where 0 <f < ae + u() ax i+  u~x a2 ) . u i Ax i1 u i+ a4 u() ax u(a) ax (4. 4. the convection is dominant over the diffusion. 4. 4.58 Eq.28) Ax (4.4 and Eq. Pe also goes to zero resulting in no diffusion error in Eq. the diffusion is dominant over the convection.30) Both Eq. Therefore. the profile of 4 will be linear and the truncation error will be reduced to zero. 4.29 from Eq.5 may be approximated as follows.e.26. i.e. i.28.31) ND .24 and Eq. If the Peclet number is small..28 and Eq. the truncation error diffusion constant for a convection dominant problem may be given approximately as follows. D uAX 2 (4. It can be shown that as P goes to zero.
2 u + v 2 + +( ) (4. two dimensional.32 is finitely differenced using donor cell scheme for the convection term. 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. A steadystate. Since the gradient of 4 in the flow direction is negligible in a convection dominant problem. the total diffusion quantity of 4 due to the truncation error is not significant in comparison with that due to the crossflow diffusion.1. conservation equation with no source term is given in the following.59 4.32) = x 3y Equation 4. u i 6x ilj + v i by ij_a(i+1 2 4x i + il1 +ij+l 2i 2 + 0iji by (4. one is the truncation error diffusion and the other is the crossflow diffusion.3 Truncation Error Diffusion in a Two Dimensional Problem There are two sources of numerical diffusion in a two dimensional problem.33) . 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.
then the truncation error of the convection term is given as follows.j+ + v(a) ly i+h.1.34) where tane = Ay/Ax tane1 = v/u The coordinate (x.35) .n) ( = x cos6 + y sine I = x sine + y cose (4.x[cos + (cose tane 1 + sine) + sine tane ] ax xY (4. transformed to the coordinate ( .y) is by rotation.v( ay ) ij . 4. ij Ax 1 1x Axhy 'i1 i Ay iji 1 +AxAy fj+ j i+ i axx dt + Yj+ yj+ yj xi+4 xi+ a (v)dx dy ay Xiih + u(t) x i+ a =U(ax) )3x .60 The approach in section 4.2 is applied to Eq.33.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 = .
37) Equation 4. 4. 4. 2 2 2 + a = ( 2 ) (4.61 where 5 is the coordinate in the flow direction.34 will be transformed to the coordinate system (t.35.32 is reduced to the following form.36) where U = /u2v 2 Now Eq. 2 (A)  [(sine tane1 + cose) 2 + a2 (sine + cosetan6 1 ) a + (0) 4 an U" (4. y v e Sx Then Eq. e is equal to 0 or 7/2.37 shows that the diffusion component in the n direction (normal to the flow direction) is zero and the cross differential term also vanishes when e is equal to e1 Therefore. dominant truncation error diffusion occurs When only in the flow direction.n) given by Eq. the problem is basically one dimensional and the truncation . 4.
that will be named here as the crossflow diffusion.62 error diffusion occurs in the flow direction as in a one dimensional problem. [67. The crossflow diffusion comes from the multidimensionality of the problem. The origin of the crossflow diffusion is illustrated in Fig. 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 4.1(A) is transformed into .3. tional false diffusion was explained by Patankar [52] and Stubley et al.4.l1(A). the hot and cold fluid enters the mesh from the left and bottom surfaces at 45* angle.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. The hot fluid will come out of the top surface and the cold fluid out of the right surface. 4. which shows a single mesh with pure convection.1. therefore it exists only in two or three dimensional problems when the flow direction is not aligned with the mesh configuration. In the donor cell scheme Fig.68) and it was clarified that this is the dominant source of the error in most multidimensional problems. It is entirely different Recently this addi from the truncation error diffusion. In Fig. 4. 4. 4.
63 Hot Mledium Hot 7 Hot Cold (A) Cold (B) Cold (C) Fig. 4. Illustration of the crossflow diffusion in a single mesh with pure convection .1.
&x u v ax Ay The effective diffusion constant for the crossflow diffusion will be calculated in the two dimensional Cartesian coordinate in Fig. Another illustration of the crossflow diffusion is given in Fig.and intermediate temperature fluid will come out of the top and right surfaces. 4. of the value of 4 causes the crossflow diffusion. 4.3.1(B) where the velocity components normal to the surface are considered.1(B) are again interpreted as Fig. 2 (lp)[l+p+p + . and the sum of the values This distribution along the points A1 . ] = (lp)4 = where u . . . at those points is exactly equal to 4. A 2 .64 Fig. . Then homogeneous mixing occurs in the mesh. CAB which looks like a long onedimensional rod.2 which shows a row of meshes aligned in the xdirection. . in The numerical results Fig. diffusion occurs in multidimensional donor cell differencing of the convection term. 4. This is a pure convection problem Consider the fluid element The fluid with no physical diffusion.1(C) by a Therefore. 4. In the donor cell scheme it is distributed all . The value of 0 at the point A should reappear at the point X because there is no physical diffusion. an appreciable amount of false program user. 4. A 3 .
Illustration of the crossflow diffusion in a row of meshes aligned in the xdirection with pure convection .2.65 (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.
AY B Ax t^ t where t t0 = tane = v/u u/Ax v P U = 3/U tane= Ly/Ax Ax by Fig. Geometry for explanation of the crossflow diffusion in a two dimensional Cartesian coordinate . V A2 e A.66 4.3. 4.
and is translated The value 4A at the point A should reappear at the Point A'. the following right hand side diffusion constant results. Gradient) A /k 1 =A of 4 Current 2 2U of 4 Therefore. The diffusion occurs both to the The diffusion to the right hand side is considered first in the following. In the donor cell scheme it is divided into two portions. p4A and (lp)4A and appears at the points Al and A 2 . right and left hand sides. Diffusion A t1 .67 element was on the line CAB at time t in the velocity direction. sine cose cos= sx Uacos61 sin(O+e ) I DRHSS . v. The travel of the fluid element is analyzed in Fig.t 0 PA £3 (Current) 2 1 constant ) where 1 = (Gradient) = U3 Lx/cose.4. which shows that the value 4A diffuses out along the fluid element. 4. k2 = Ax sine/sin(el+8) k3 = Ax sine 1 /sin(e 1 +8) After substituting appropriate expression for each term.
Ya. 4. Geometry for explanation of the crossflow diffusion in a rotated two dimensional Cartesian coordinate .ct t2 t3 Fig.4..68 ti .
D D = uAx(lp) = VAy p (4. There are two diffusion components in a two dimensional problem as follows.40) .38 is the effective diffusion constant when the diffusion is confined to the one dimensional fluid element. 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. sine cose RHS LHS cos1 sin(+8) (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.39) Equation 4.38) Equation 4. The crossflow diffusion constants are the ratios of the current and gradient of 4 in the x and y directions.39 is identical to the following simple expressions. 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.
D DDM = UAlxy sin2e= 3 3 4(by sin36 + Ax cos36) UAx tan6 1 sin26 3 4(tan8 sin6 3 + cos3) (4. < 6 < 81 or < . 4.38 should be differentiated with respect to 6 in order to find out the velocity direction with maximum crossflow diffusion. Now the analysis will be extended to a three dimensional case in Fig. De Vahl Davis and Mallinson [24] also derived the effective crossflow diffusion constant and their result is given in Eq. RHS D = It can be shown that. 4. In Fig.42.41) Equation 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. 4.42) Equation 4.70 where p = u Ax u v By Equation 4. 4. In the donor cell p' y scheme the value 4 is divided into three portions px 0 . Eq. 4.42 underestimates the crossflow diffusion as shown in Fig. (4.41 shows that DRH S (or DLH S ) is maximum at an Tr7LHS angle of 61 < 6 e between 4 and e1.42 give the same result. that is to say.5. When both 6 and 61 are equal to . 4. 3 0 when tan 6 = tan6 1 .40 and Eq.6.
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. 4.71 Cross flow Diffusion 2.5 80 \ % % 4 1. 0.0 0 0 .5 I I 0.0 150 300 450 600 750 900 Fig.
4.6. Geometry for explanation of the crossflow diffusion in a three dimensional Cartesian coordinate . p Fig.72 z C V Az X 0 o a X AY Ay B y x z C p .
B. n C denote The coordinate of the Any point on the plane ABC should satisfy the following Eq. £4 cosy) nB . S= 4 + y z zAx 1 cos+ + cos Ay Ax + cosy Az (4. along the directions XA.and nA unit vectors along those directions. point X is given as follows. 4. t4 cos8.43.73 z occurring at the points A.43) Since the point X is on the plane ABC. XB. cosa cosB cosy (4. CX which are denoted as £i' z2' k3 can be obtained as follows.44) 4 Therefore. the following relation holds. S+ Ly Ax + Az = 1 plane ABC (4.45) The lengths of AX. This is the socalled crossflow diffusion phenoThe currents due to the crossflow diffusion are . menon. and C instead of at the point X. XC. X(£ 4cosa. BX. £1 = AX =/(x4cosa)2 + (4cos)2 4cos) 2 = Ax 2 + 42 2Ax£ 4 cosa .
(Az 4 cosy)] (4. z directions. 2 o C 14cosy] = 1  £ 4 cosa.OX (4. nB' n C can simply be given as the products of the transported quantities and velocities.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. Z 4 cos8. Therefore. y.  n A = XA +1A  OA .47) etc. I (Current)xAi XAI = px Px C1 1 Lt . n C can be obtained in terms of their components in x. 4 cosy] 1 nB = [i4cosa. nB. (Ay4 cosS).48) The currents in the directions nA .46) + The unit vectors nA .
(Grad)x = 4. y./ Ax (Grad) = 4.51) p Ax D x = lt (Axk 4 cosa) + p Ax . z directions are given as follows./ Ay 4/6x Q/by (Grad)z = 4/Azl (4.50) Now the x direction crossflow diffusion constant may be given as the ratio of the current to the gradient of 4 in the x direction. Px D (Ax£ + Py D + nC(4 (Current)x 4 cosa) c osa) At ( 4 (cosa) (Grad)x = 4/Ax Therefore.49) The gradients of 4 in the x.(4 pz x co s a ) ct 4 + at ( 4 cosa) 4 .75 I(Current) XB = Py p2 t I(Current) XCI where At = k /U z '3 At (4. (4.
most important ones may be the skew differencing. p Ax D = ux[ £4cos x . finite element method.53) The crossflow diffusion constants in the y and z directions can be obtained in the same way. tensor viscosity method. indices will also give the same result.52) Inserting the expression for At given by At = £4 /U.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. although none of them has been successful enough to be accepted widely. . Dy = vLy(lpy) Dz = wAz(lpz ) The rotation of (4. This section is a review of the schemes that The have been used to eliminate the numerical diffusion. we can simplify the result as follows.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. etc.54) 4.1 = uAx(lpx) (4.
.. Basically the inclusion of four additional corner points contributes tomore accurate numerical modelling of the convection.II1~LI1I~~ ICII1_1___~1___*_____11__ 77 Most of the schemes in Table 4. but directly along the velocity direction. but just appreciably lower than that of the donor cell scheme. In the skew differencing scheme the 5point relationship becomes a 9point relationship. skew differencing and corrective schemes..4 fall in the category of a corrective scheme. The crossflow diffusion is not eliminated entirely in the skew differencing scheme. the interpolation should be done between neighboring points to obtain the value at the upstream point.3..^ILli. The corrective scheme is to reduce the diffusion constants in order to compensate for the additional false . In the skew differencing type schemes finite differencing of the convection term is done not in terms of the x and y directions. the interpolation between neighboring points may still give some crossflow diffusion. The donor cell treatment of the convection term results in a 5point relationship in a two dimensional problem.1 can be categorized into two types. Since the upstream point in the velocity direc tion does not necessarily fall on the mesh point where the quantity under consideration is defined. However.. 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. S.~_ 1 i_.
1981 [65) .aptjve me.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 !.ronLUD Lillinion/VUDCC Nuana.j.1QUICK t ETovh ci . Wilkes Elbahu ct l.78 Chss' iction Cen:ered .thod of chaaceriics Fiite anuryic method Tenso.ad. Esposito Glas and Rodi Huffenus Lnd Khali./LLUE Huint/QUICK Wilkes/LUE Lillimpon/SUD LiiL.1.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 . upwind f~iite di Teretn atl.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 .Highe cdtr upwi:nd :"'ie Cdi't.ordc.e elements Clifft Dontcl t al..hod Mesh *txLnsformation rr. viscosity method Sel.inte Conribo:r/me. Grandotto Hickmott Firs.thod LLxW'en doff  21 x11 80 x 400 Table 4. List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research. £Ebzhau t aL/PATANKAR Lilinpion Orlandi Priddin Wada cit l.
55) where n is the distance normal to the flow and U = /u 2 +v 2 . stants in x.79 diffusion.54. where the balance equation is set up for a given control volume by considering convection and diffusion at the control surface.53 and Eq. The corrective scheme is valid only if the effective numerical diffusion constant can be predicted accurately. D.54. Since the crossflow diffusion is dominant over the truncation error diffusion and the effective diffusion constants for the crossflow diffusion can be predicted theoretically. The correction formula The diffusion con used here is Eq.53 and Eq. 4. 4. 4. The tensor viscosity method is one example of the corrective scheme. 4. 4 = C1 + C2 n = C1 + C2 (  (4. The finite difference equation for Raithby's scheme is derived on the assumption of a linear profile of 0 normal to the flow and a uniform profile in the flow direction. the corrective scheme can give a numerical solution which is almost free from numerical diffusion.3.1 Raithby's Scheme Raithby's scheme [53] is an Eulerian type of approach. It is therefore a conservative scheme. y and z directions will be reduced by the amounts of D .. and D z in Eq. although the validity of its correction formula is not clear. 4.
control volume and notations in Raithby's scheme . the calculation domain showing grid lines.7. A segment of.80 Fig. 4.
This should be repeated for the four control surfaces of every mesh in a two dimensional problem. u 4 + 'v.= U ax dy SC (4. it is much simpler than Raithby's scheme and becomes conservative in a unidirectional flow. 'ij US (4. 4. The convection term is differ enced in the velocity direction between mesh points. which may be too complicated and time consuming for a practical purpose.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.56) where U = /u is the coordinate in the velocity direction and 2 +v 2 .57) where the upstream value €US may be calculated by .2 Skew Differencing By Huh Skew differencing scheme here treats the convection term in a Lagrangian way. Although this scheme is nonconservative.3. The inclusion of the four additional points also complicates the structure of the coefficient matrix in an implicit scheme. not in terms of the convection that occurs at mesh interfaces.
3.2. uv (4. The validity of this scheme is based on the fact that the crossflow diffusion can be predicted accurately. Dx = uAx(lpx) D Yy = Ay (lpy ) The expressions for .4 Corrective Scheme By Huh The corrective scheme is to use the reduced diffusion constants. 4. DD y .53 and Eq. 4.58) t2U= Eq. 4.3.58 for the numerical diffusion constant is apparently not consistent with the results in Eq.82 interpolating two neighboring points. the crossflow diffusion constants D in section 4.2.3 Tensor Viscosity Method The tensor viscosity method treats the numerical dif fusion constant as a tensor quantity. D .3. Some of the results of this skew differencing scheme is given in section 5. z directions to compenx sate for the crossflow diffusion effect. DDx. DD z in x. 4. D are derived . 4. y. the tensor viscosity [25) T = Lt uu u2 The expression for is given in the following.54.
Dz at the center of a mesh where all quantities are defined except the flow field.1 Mesh Point Implementation The mesh point implementation is to assign the crossflow diffusion constants Dx . Dy . The velocities at two boundary faces are averaged to get the representative velocity. In a coarse mesh or recirculating flow field the latter implementation strategy is definitely preferred._IIU__lpULI~~I___~lLlt i~Lli ~i ~iL~pn~sl.. . the mesh point implementation and mesh interface implementation.+ w Ay Lz x etc.These two become identical when the flow field is uniform or a very fine mesh is used so that the flow change over neighboring meshes is negligible. After the crossflow diffusion constants are assigned to every mesh point.4. which is used with the cell dimensions to get the crossflow diffusion constants.3.59) v u . 83 D z = wAz(lp z ) where px = u (4.+ . Two implementation strategies have been tested so far. they should be averaged again between neighboring mesh points to calculate the crossflow diffusion current at mesh interfaces. 4.
4. The mesh interface implementation is always recommended although it has a more complex program logic than the mesh point implementation. 4. The same logic can easily Appendix C includes be extended to a three dimensional case. U and V. 4. . The two velocity sets (Ul. so that they are proportional to V1 and V2.3. Dyl and Dy 2 . and the resulting D x and D y will be assigned to each interface.8. can be directly used with cell dimensions in Eq.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.Vl) and (U2. For the second case the two outward velocities. There are three possible cases that should be treated independently in a two dimensional case. U1 and U2.V2) are used to give (DxlDyl) and(Dx 2 'Dy2 ). as shown in Fig. The right interface will have the crossflow diffusion constant (Dx1+Dx2 ) and the top and bottom interfaces.84 4. The third case is expected to have a flow split as shown in the figure. 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 velocity U is split into two components. For the first two cases the cross flow diffusion constants are zero because the velocity vector is aligned with the mesh configuration.59.
8._~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 .
3. for the general conservation equation.62) . which is a parabolic partial differential equation. Implicit and ADI (Alternate Direction Implicit). conservation physical constraint accuracy stability 2 ax 2 2 y 2 +ux +at + 7 + T7) + S sz (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 ) (4.86 the fortran program for calculating crossflow diffusion constants in an arbitrary flow field. 4. Explicit.60) Explicit (lD): n+l I n i At + u (i i x n i. The pros and cons of the three schemes will be compared in the following four viewpoints. Eq. 4.4 Comparison of Explicit.60. 4. 1.) = i1 n D Ax ( i+1 2i l n n + n ) + iS i1 (4. Implicit and ADP schemes There are three typical solution schemes. 2.
63) 4. when the net inflow or outflow at . 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 24.4.87 ADI (2D): n+ ij (A) _n ij In+ + u ij _ cn+ in + v n ij1 At 2 Ax Ay )n+ i+lj 24 n+ Ax 2 +n+ n l + D j+l n ij 2 Ay n+l ij n iji + S n+l ij n+ ij n+ ) ij n+ i1 n+l ij1 At 2 Ax Ay n+ =D i+i S2. + 4).. or affected only by the source term contribution (SO). 2 Ax n+l 2 * ij v n+l ) ij n+l Ay 13j1 n+1 ij+l n+l 1j1 At Ay2 + S (4.1 Conservation The conservation principle is that the total amount of 4 should be conserved (S=0).
A scheme may be both stable and nonconservative and also both unstable and conservative because the conservative property is independent of the stability.4. Although a conservative form is preferred to a nonconservative one. implicit and ADI schemes can be made conservative by using appropriate differencing forms for the convection term. In order to guarantee conservation.88 the domain boundary is equal to zero. the conservative property is as important as the stability because the error from nonconservation may accumulate over many time steps. When we are concerned with a steady state solution. there are some nonconservative solution schemes like ICE (Implicit Continuous Eulerian). All the explicit. + CI 4. For a long time transient. Although they usually give . the conservative property may not be crucial.2 Physical Constraint The physical constraint is closely related with the stability in a practical sense. where the convection terms of momentum equations are linearized in a nonconservative way. the same exchange term should be used for neighboring meshes. however.
The Courant condition and Hirt's stability condition [36] that the finite domain of influence should at least include the continuum domain of influence are good examples of the physical constraint condition. 4. )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. R C AxAy = CAxy + D R Ax Ax CAy C y At At ¢4AxAy = 1RAxAy .89 similar constraints on the time step size and mesh spacing.2.4.1 Diffusion There is a maximum net flow that can occur between neighboring meshes by diffusion. they are from entirely different origins.D . The laws of physics impose some restrictions on the numerical scheme to get physically reasonable solutions. new values of 4 are given as follows.
t' > Q' .64 or 4.65 is used. . we have the following physical constraint. d 2D t < 1 (4.63 is unconditionally stable. should still hold. however. 4.65) Ax Equation 4. scheme given in Eq.90 Since the relationship.64) Eq. the solution will be unstable or stable with damping oscillations.65 may be a too conservative criterion in a real problem.64 may be extended to a ddimensional case as follows. 4. Therefore the physical constraint should be respected in addition to the stability condition to get a meaningful transient solution. Both of the oscillation and instaThe ADI bility should be avoided in a transient problem. When the physical constraint is ignored and the time step size greater than that given by Eq.64 or 4. 4. there will be a damping oscillation in the solution when the time step size is greater than that given by Eq.65. 4. R AxAy(4) C) > 2D (R Ax CyAt 2 DAt X2 < 1  (4.
a portion of the control volume C will cross the interface S. At < 'x u (4.4. Therefore. the damping oscillation or instability may occur as in the case of diffusion.2 Convection C A u B C A ). than the Courant limit.66) When a time step size greater than the Courant limit is used.4. Then uAttay is no more an appropriate expression for the convection quantity during At.2. 4.91 4.B 4y S The explicit and donor cell treatment of the convection term gives the convection quantity at interface S as ulAAtAy. It is based on the assumption that 4A is the repreIf At is greater sentative value of the control volume A.3 Accuracy Realistic interpretation of a finite difference equa tion comes from an integral form of the conservation .
4. I I t i r _ I I [ n n+1 n n+l _ 0 It Explicit Implicit CrankNicholson $i+. Profile assumptions of Q in space and time for various differencing schemes .9.92 n+l n+l 4' I 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.
4 Stability The stability is a mathematical problem. However. 4. If its absolute magnitude is less than one. 0(Ax 2 ) is The also not appropriate for finitely large At and Ax. It depends on the maximum eigenvalue of the iteration matrix. The stability can be checked by the theorems about matrix eigenvalues or Von Neumann analysis.4. the accuracy depends on the profiles 4 in time and space because the exact profiles assumed for are not known a priori. the error propagation is suppressed and the largest error contribution will be from the previous time step. accuracy depends on how well we can compute the convection and diffusion quantities at control surfaces over the time At. the accuracy is usually the least important among the given four considerations. whether the error introduced at a certain step will increase or decrease as the calculation goes over to the following steps. The accuracy expressed in terms of O(At). The Von Neumann . Since the exact profile depends on the Peclet and Strouhal numbers of the governing equation.93 equation. 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. In other words. The infinitesimal approximation of differential terms is not appropriate in most problems of our consideration.
2 2i + D i+1 i n 2 i i1 (4.. 1 4n = E~nei where I = .1. lowing sections.4.4. (4.67) The solution 4. The stability is not a sufficient conIn the fol dition for an acceptable numerical scheme.4. n+l i n €i+i I + u i+1 n ii 2Ax i_1 lt n n i+l .68) .1 Explicit Scheme The Von Neumann stability analysis of a finite difference equation is simple in its idea. implicit and ADI schemes using the Von Neumann analysis. is expanded in Fourier series as follows. 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.94 analysis is to expand the solution of a finite difference equation in Fourier series and check the decay of each mode separately. but its algebra may get complicated.4. stability criteria will be derived for the explicit. 4.
the function .68 is substituted results.95 The absolute value of should be less than one for every When Eq.70) Equation 4.71) The function f(t) will be defined as follows. 4.67. f(t) = i 12 = (4d2_ 2 )t2 + (Sd 2+4d )t + C 2+4d 24d (4. Therefore.2dx(t) where t = cose 1 < t < 1 ] 2 +C 2(1t 2) 1 (4.72) In order to get the stability of Eq. the following expression for = lCxsin xI where cx d x uAt A tx  2d (lcos x) (4. 4.2dx(cosex) 2 + [Cxsinex 2 < 1 (4.67. 4. 2 11 = [1 . in Eq.69) Dbt Dt 2 Tx. [i .70 may be rewritten as follows by defining cos6 x as t. mode to achieve stability.
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.73) for any t in [1. D(Discriminant) = (C 2 .74) One of the two meeting points with the taxis is the point t=l because f(l) is equal to zero. The parabola f(t) always meets the taxis because the discriminant is always greater than or equal to zero.73 can be solved by a graphical method using the characteristics of a quadratic equation.75) The parabola f(t) may have three distinct shapes according to the sign of the coefficient of the second order term. Therefore. f(t) < 0 (4. f(1) = 0 (4. 4.1] Equation 4. .96 f(t) should be less than or equal to zero for any t in [1. (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.2d )2 > 0 (4.10(1).1].
10. Three distinct shapes of the parabola. according to the sign of the coefficient of second order term .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... 4. f(t).
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. 4.77. 4d 2 C 2 4d C > 0 < 0 : : 0 < d dx > < 2 Cx The stable region in the coordinate system (Cx. 4.76 and Eq.2 One Dimensional Donor Cell Differencing of Convection Term The following Eq.C (4.10(3). 4.4. 4. 4d 2 . t=l as shown in Fig.78 is a finite difference equation with donor cell differencing of the convection term.1.98 f(l) = 8dx(2dx1) < 0 0 < d < (4.77) Summing up the results in Eq.2d Axis of symmetry: t = 4dx dx > C 2 x > . 4.11. we get the following stability conditions. 4. .dx) is shown in Fig.76) In case (2) f(t) is a straight line and f(l) again should be less than or equal to zero.
99 d x 1 1 0 1 dx = C2 x 2x C Fig. Stability ondition in the plane (Cx.d ) for an explicit scheme with central differencing of convection term in a one dimensional problem . 4.11.
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)
1 SRe 2 al +2 1+a +a 2 Fig. . 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 .104 Im \ .13. 4.
should exist for the stability of an explicit scheme with donor convection term in cell differencing of a general two dimensional problem ..105 .14. . The region in the complex plane where the amplificatior factor. 4. Fig. .Re Both al and a 2 should be in the shaded circle to ensure that Ial+a 2 +1I< i.
15.dx) and (C y. which is independent of the time step size.16. 4. When Ax and At are already given.89 and al and a2 may be rewritten as follows. The stability condition is again given by Eq. which depend on both the mesh spacing.89 can be solved by the graphical method as in the previous sections. 4.e Rx . 4.88 and Eq. 4. At.4.1. the stable time step size for a given mesh spacing. Therefore. and time step. the stable range of Cx should be determined in terms of the cell Reynolds number Cx/dx'. where the points (Cx.88 and 4. the stability However. 4. The results are given in Fig. Cx and d .106 Both Eq.91) . our usual concern is can be checked directly.90) a 2 = C [1 y R R Y + R (4. A three dimensional case can also be solved through the same procedures and the results are given in Fig. The stable region shrinks on a linear scale with dimensionality. Ax.d) should be in the shaded region for stability.e Y) (4.4 Stability Condition in Terms of Cell Reynolds Number The stability conditions in the previous three sections are in terms of the dimensionless numbers. I 1 2 + (1+)e Rx Rx 2 + (1+)e a1 = Cx [1 + Ie 1 3 .4. At. 4.
Stability cc. 4.107 d (d ) d = x 2 1 x + 1 4 C x(C ) 16 d= C2_ C x 2x x Fig.15.dy) fcr an explicit scheme with donor cell differencing of convection term in a general two dimensional problem.idition in the planes (Cx. .dx) and (Cy.
108 d xy(d or d z ) Cx d dx + 1 1 Cx (C d! S3 2 .(C Cx x y or Cz) z d ) Fig. Stability condition in the planes (Cx.dx) etc. 4. for an explicit scheme with donor cell differconvection term in a general three encing of dimensional problem .16.
__LI___IIII__LI*X__ ID I 109 The following condition for Cx and Rx can be derived from Eq.88 and Eq. 4.1] f(l) = 0 (4.92 can be simplified as follows.C (+2x . [ . 4.94 ) D(Discriminant) = C 2 (K2C ) 2 > 0 (4.95 ) The function f(t) satisfies Eq. 4.93) ) where 2 K = 1 +. and has distinct shapes for the Following three cases. 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 [1.94 and Eq.) RX (1cose )) x + [C sine] x x < (4. f(t) = C (1+K) 2 (lt) . (1) Cx 2 K2 1) > 0 concave upward .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 (4.92) Equation 4.95.90.
yu < 1 (4. The stability condition in Fig.17. 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. although it can be relaxed if there is a constraint on the velocity direction and mesh spacings such that the following' relation is satisfied.17.98) Ax . to zero.97) The results in Eq. (4. V f . The same result will be obtained for the ydirection.110 (2) C 2(K 21) = 0 linear (3) 22 C 2(K 2_1) < 0 concave downward For case (1) and (2). R > 0 x : 0 < C < x 2 2 2(1+) x and (4. therefore the subscript x may be replaced with the subscript y in Fig. 4.96) R 1 x < 0 < C x < (1+2) Rx . 4.96 and Eq. solution procedure is similar to those in the previous sections and the results are given in the following as.17 is applicable to all general 2D problems.97 are shown in Fig. 4. t=l. 4. 4.
Cy) for an explicit scheme with donor cell ditferencing of convection term in a general two dimensional problem . 4. Stability condition in the plane (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.Cx) and (Ry.
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.18.Cx) for an explicit scheme with donor cell differencing of convection term in a two dimensional problem with f v u/Lx .
4.19 and Fig. 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.C Y y . (4.113 Then the number replaced with on the C axis in Fig.17 and numerical experiments.99) (4. d >  greater Cx (4.99) The stability condition is shown in Fig. 4.100) For a general two dimensional problem.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. and it can be seen that the implicit scheme is unconditionally stable if the diffusion constant is than . 4.20.uLx.4. the stability condition is found to be the following.18.102) dx > .2 shows the com 1 l+f in Fig. 4.Cx d > .101) (4. 4.2 Implicit Scheme The following Eq. 4. parison between the analytical result in Fig.17 should be Table 4.4.
720 0.403 0.833 8.333 833.180 0.333 83.780 0.488 0.2. R x 0.167 2.120 0.660 0.019 Numerical Experiments 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 .333 4.083 Von Neumann Analysi s 0.114 Maximum Stable C.147 0.120 0.629 2.260 0.3 8.499 0.540 0.060 Table 4.660 0.380 0.
115 d =  x d =.dx) for an implicit scheme with donor cell differencing of convection term in a one dimensional problem .19. Stability cndition in the plane (Cx.C x 2x x 2x C x dx 1(C2+C 2xx ) Fig. 4.
20.Cx) for an implicit scheme with donor cell differencing convection term in a one dimensional of problem . 4.116 Cx=1  2 Rx 2 1 0 R x R C x R +2 Fig. Stability condition in the plane (Rx.
jto be less than or equal to one.4. 2 _C C 2 {[(cos )(+d X))] C [sine I2} {(1cos6 C 2 )(. One of the formulations has both the x and y direction components in each step.63. The Von Neumann analysis of the first formulation in Eq. 4.63 gives the following stability condition.117 4.104) .3 ADI Scheme There are two possible formulations of the Alternate Direction Implicit (ADI) scheme given in Eq.103) jsj <1 In order for I.+d )] C ) {[1+(lcosex +dx)) +[fsinex }{[l+(1cosy) (+dy) + Y[siney } (4.4. 4. the following condition should be satisfied. C d x>2 (4. treating them implicitly one by one at each fractional step and the other formulation has only the x or y direction component treated implicitly in each step.
105) The stability condition of the second formulation in Eq. . C d x > 2 2x (4.106) C d y  >  ~ 2 (4.107) Therefore the second formulation of the ADI scheme in Eq. C d y>  2  4 (4.118 The same condition should hold for the ydirection. 4. 4.63 can also be given by the Von Neumann analysis as follows.63 has the same stability condition as the fully implicit scheme.
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. occurs as a result of the buoyancy force due to density .1. 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.2. 5.1 and Fig. 5.1 Natural Convection Results The containment is modelled as a two dimensional rectangular compartment with an obstacle as a heat sink. For example. Finally. density is determined by temperaSince the natural convection of air ture and pressure. Natural convection occurs due to the heat transfer between the obstacle and containment air. 5. the ADI scheme is explored to use a time step size larger than the Courant limit to save the overall computational time. energy convection and heat transfer models are considered and the resulting flow fields are given in Fig. 5.
0 Btu/lbm 0F (Q ob = 208. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 10.1.4 sec AT = 100F T .8537x10tu/ft20 F sec (Cp) = 1.120 p ). = 677. 0.29 Ibm A= 2 ft2 T. 5. = 680F air T = 580F Fig.8 sec .5 ft/sec h = 1.
29 ibm (K 2 A= 2 ft T. 5. 5 ft/sec h = 1. T ab = 580 F Fig.4 sec 1 0 AT = 10 F T air = 680F .5 sec .0 Btu/lbm OF (c )b = 208.121 > 0. = 677.8537x10 tu/ft2 F sec p)ob = 1.2. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 62.
The equation of state of an ideal gas is given in the following.total Pold Told new old Qcell cv Mf. not on the pressure because of a high modulus of elasticity.1) T Pnew = old + Ttotal c v M f. time Pressure of the fluid at old and new steps. new Pnew RT new (5. temperature and pressure are equally important parameters in determining the flow field. c : Specific heat of the fluid at constant volume. Pold' Pnew: . depends primarily on the temperature. The density change of liquid.cell where Pold' Pnew: Density of the fluid at old and new time steps. Told' Tnew: Temperature of the fluid at old and new time steps. however.122 gradients.
2 in terms of the dimensionless numbers.4 when internal energy is used for enthalpy in air flow. 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.1.1. .cell. An experimental correlation is given in Eq. The energy convection will be underestimated by a factor of 1. k = c /C for liquid and 1.. . diatomic and polyTherefore.total : Mass of the fluid in one cell and in the total domain.2 Energy Convection Another difference between liquid and gas is the ratio of the Specific heats. The Nussett and Grashof numbers here are based on the axial length of one computational mesh. 5. but they may also be based on the vertical height of the obstacle. The value of k is 1.123 Mf.0 Monatomic.4 for air.3 Heat Transfer Modelling The heat transfer to the obstacle is modelled as a natural convection from a vertical surface. the atomic gases have different values of k. convection energy transfer should be in terms of enthalpy instead of internal energy. Mf. Qcell total : Heat input to the fluid in one cell and in the total domain. 5. 5.
923 x 103 Therefore.021 n 0.25 0.sec .2 ft/sec' 5 V = 1. Gr = 9.3) Pr = 0.2179 x 10 B = 1.4 lbm/ftsec Btu/ft 2 F.124 Nu = C(GrPr) n ( 5.109 > 109 C 0.0763 ibm/ft 3 g = 32.7134 x 108 Nu = 89. This may be a typical value for the natural convection heat transfer in the containment without any phase change.4 The following data are used to calculate the heat transfer coefficient at the obstacle wall.876 The heat transfer coefficient is given as.555 0. L = 2 ft AT = 500 F p = 0.708 (air) GrPr 105 . . h = 1.8537 x 10 .2) 3 2 Gr = gS(AT)L p 2 (5.
4 sec.5) where M A : Mass of one computational mesh. can be obtained from heat balance between the obstacle and fluid as follows. 5. T. Therefore. CQ = Cpob (TT) (5. (Mc )ob (MC) obf hApo b ob f (5. If the flow is turbulent.1 and Fig.4 Turbulence Modelling Laminar flow is assumed in obtaining the results in Fig. The time constant.125 The VARR input [21) requires the time constant of the heat transfer instead of the heat transfer coefficient. Specific heat at constant pressure. pf Pob: c h : Fluid density. 5. the heat .4) where CQ is the source term in the energy equation of fluid.2. Obstacle density. T = 677. Heat transfer area between the obstacle and fluid over one computational mesh. 5.1. Heat transfer coefficient between the fluid and obstacle.
The corrective scheme is tested in recirculating flow problems for mesh point and mesh interface implementations. 5.7 show that excessive numerical diffusion has occurred due to the . 5. steadystate problem in Fig. The flow in Fig. is compared with De Vahl Davis and Mallinson's in predicting the magnitude of crossflow diffusion.4 and Fig. 5. 5.2.5 show that the numerical solutions is close to the analytical solution and the truncation error can be neglected. 5. Figure 5.3.126 transfer. 5. The results in Fig.3(A) is parallel with the mesh orientation so that no crossflow diffusion occurs. 5. 5.to the obstacle will be enhanced and the diffusion transport of momentum and energy will also be enhanced in the containment air.3(B) is skewed at 450 to the mesh orientation and both truncation error and crossflow diffusions are included in the solution error. The skew differencing and corrective Huh's formula schemes are tested for various problems.2 Numerical Diffusion The truncation and crossflow diffusion errors are compared in a simple geometry where an analytical solution can be obtained.1 Truncation Error and Crossflow Diffusion The truncation error diffusion is compared with the crossflow diffusion in a two dimensional.6 and Fig. The flow in Fig.
0 BT S=0.0.0 a= 1. 5.0 u = v = 7.0 (A) (B) AB=BC=5. for evaluation of numerical diffusion .0 ay AB=BC=5. Problem geometry for the two cases.0 B T specified Fig. (A) parallel flow and (B) cross flow.0 x BT  @x a( aT + 2 2 a2 ay ) Th= 10.0 B T_ 0.3.07 (B) =0. 0 ST By =0.0 u = 10.127 (A) .
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. 5. 5. spatial coordinate Fig.128 Analytic solution Numerical solution with Donor Cell T 10.
3(A) .~ 129 Analytic solution Numerical solution with Donor Cell C spatial coordinate Fig. Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CD in Fig.__LL l.5.__~ .~ I~. 5.^XI ill I~LX_L~LI. 5.~l r.
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.3(B) . 5. Comparison of the analytic solution.6. 5.130 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Df cf T 10 spatial A coordinate Fig.
5. 5.131 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Dcf T 10 / I / / ..J 0 I' C spatial A coordinate Fig.3(B) . . 1/ 5.3(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.7. Comparison cT the analytic solution.
6 and Fig. 5. 10 x 10 meshes and Fig. 5.54. 5. 4.2.9 is the result by Huh's formula and Fig. 5. 4.810.12 for a simple pure . Figure 5.5 and Fig. Since the corrective scheme is based on the validity of those correction formulas. 5.2. 5. it is necessary to test them for various cases.3 Skew Differencing Scheme Two skew differencing schemes. 6 = 6 = 450.4 and Fig.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. 5. solution. Eq.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. 5. more diffusion has occurred in Fig. It can be seen that the former result is much better than the latter. The problem geometry in Fig.53 and Eq. De Vahl Davis and Mallinson's and Huh's formulas give the same crossflow diffusion constants for the case. 5. 5. Figure 5. Figure 5. 5.2 Validation of Huh's Correction Formula The crossflow diffusion can be predicted by Huh's formula.8.132 crossflow diffusion. Raithby's and Huh's.11 and Fig.8 has the arbitrary angles of 6 and 61 such that 6 = 600 and 61 76.6 are for 5. are tested in Fig.7 than in Fig. 5.10 by De Vahl Davis and Mallinson's for the problem in Fig. meshes.6.
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 .0267 u = 5.8.335 Huh A e1 = 76.0 v = 8.7674 DH u h = 3.133 Yo 2 T=0 D a = 1. 5.6603 DDM= 0.0 Ax = 0.810 8 = 600 8 Meshes 24 Meshes Fig.2406 by = 1.
(D+Dcf).8 . 5 0 spatial coordinate Fig. 5.9. by Huh's formula along the line BD in Fig.134 Analytic solution with D+Dcf Numerical solution with Donor Cell T 10. 5. Comparision of the numerical solution with donor convection term and cell differencing of analytic solution with increased diffusion constant.
5. Comparision of the numerical solution donor cell convection term and analytic differencing of with increased diffusion constant.8 . 5.10.135 Analytic solution with D+DDM Numerical solution with Donor Cell T 101 spatial coordinate Fig. by De Vahl Davis and Mallinson's formula along the line BD in Fig. solution (D+DDM).
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 34.75 25 U 12. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with e = 45* . 5. Comparison of the true solution and solutions by donor cell scheme.63 87.25 18.25 65.38 50 31.136 Donor cell 100 100 100 100 100 93.75 81.25 0 Skew differencing by Huh (True solution) 100 100 100 100 50 0 0 0.5 0 6. 0 100 100 100 100 100 50 100 50 0 50 0 0 V Fig.5 68.11.
71 20.56 1.13 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 6.99 10. 62.88 81.137 Donor cell 100 100 100 100 50 .A Skew differencing by Raithby 100 100 100 100 96.6 0 Skew differencing by Huh 100 90.5C 100 100 50 Fig.1 6.43 7.25 1. 5.3E 21.2! 90.26 80.12.47 23.5E 25.50 2.83 72.74 38.59.67 5.11 0 11.25 53.6.01 55.25 0 0 0 . skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure covection problem with e = 63.33 70.430 0 50 0 Comparison of the true solution and solutions by donor cell scheme.8922.E 21 0 12 True solution 3.91 31.3' 40.91 71.52 1.85 0.11 3.86 16.22 9.5 0 0 63.9 33.430 .25 37.70 4.3: 11.96 17.
5. Figure 5. Raithby's scheme is an Eulerian Therefore. It is indicated that Raithby's scheme may give unphysical results in problems with steep gradients of the quantity under consideration.14 show that the analytical solution can be reproduced by Huh's skew differencing scheme for the case.13 and Fig. former is conservative while the latter is not.430 angle and neither Raithby's nor Huh's scheme can reproduce the true solution. The details of those schemes are Figure 5. 5. The purpose of the corrective scheme is. the However.3. 8 = 1e= 45*.2. to obtain the true solution by subtracting the additional crossflow diffusion .11 has a flow given in sections 4. method while Huh's scheme is Lagrangian.12 has a flow at a 63.4 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. 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.138 convection problem. but less than inthe donor cell scheme.2. Fig. both of them are conservative in a unidirectional flow. however.3. Some crossflow diffusion has occurred in Huh's scheme.1 and 4. 5.
5. 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.3(B) . 5.13.139 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh C A spatial coordinate Fig.
14.3(B) . 5.140 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. 5.
Two implementation strategies of the corrective scheme. The corrective scheme gives a physically reasonable solution. Figure. Figure 5.20 shows the diffusion corrections of the mesh point and mesh interface implementations at each interface. 5.3(B). 5. Both of them are tested in a recirculating flow problem with pure convection. although not identical to the true solution.430. Fig. 5. 5. The full correction of the crossflow diffusion gives the true solution without any numerical diffusion error. donor cell solution and the solutions by mesh point aand mesh interface implementations. mesh point and mesh interface. Fig.21 gives the true solution. are introduced in Chapter 4. Figure 5. The corrective scheme is also tested in the problem geometry of Fig. 5.15 numerical solutions are given for a simple pure convection problem with various diffusion corrections.19 is a hypothetical 3 x 3 recirculating flow field where the inlet boundary values are specified on the left and bottom surfaces.18 show that the crossflow diffusion can be eliminated by the corrective scheme and that a finer mesh spacing always gives a better solution.141 constant from the total diffusion constant of the finite difference equation. In Fig.16 has a flow with 8 = 63. The mesh point implementation gives an unphysical solution . 5. The flow is at a 45* angle with respect to the mesh orientation.17 and Fig.
0 2.3 D=0.1 10 8.01 7. 5.13 7.0 1.5 x y Fig.29 0.71 0.29 5.75 6. Donor cell solutions with various diffusion corrections for a pure convection problem with e = 450 .99 D=0.5  3.0 8.5 0 1.0 10 7.0 2.59 7.25 0 5.5 (True Solution) 9.0 2. 0 V ________ D=0.08 5.12 5.92 5.41 u=v=l Ax=Ay=l D =D =0.71 5.88 5 4.88 10 5 0 2.142 D=0.15.
v=2 x= y=l 2 D =D = x y 3 Fig. 0 D2 3 0.56  2.16.41 0.430 u=l.04 4.96 3.59 1.10 9. 5.07 0.07 2.91 0.11 0.430 .82 0.33 1.143 D=0. True solution and donor cell solutions with and without diffusion correction for a pure convection problem with e = 63.93 10 5.11 True Solution V 63.37 3.11 6.73 10 7.07 10 3.
Comparison of the analytic solution.144 Analytic solution Numerical solution with donor cell Solution by corrective scheme 10 A spatial coordinate Fig.17.3(B) . numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig. 5. 5.
5.145 Analytic solution Numerical solution with donor Solution by corrective scheme 10 5 0 spatial coordinate Fig.3(B) . 5.18. 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.
Simple recirculating flow field for test of the implementation strategies of the corrective scheme .146 0 Ofr 3 ~LC~' / 10 1 1 5 0 1 3 3.19. 3 ) 2 3 2 Lx = Ay = 1 Fig. 5.
20.25 1.75 0.75 0 0 0.125 0.2 0 0 0 mesh interface implementation Fig. 5.2 0.5 0 0 5 0 0 . 5. 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.75 1.5 0.19 .147 0.75 1.25 mesh point implementation 0.625 0 .625 0.75 1.
94 9.18 7.83 6.08 9.86 12. 5.82 8.26 8.21 7.19 .21. Comparison of the true solution.10 8.51 10 5 0 Fig.148 True solution Donor cell mesh point implementation 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 Fig.21 mesh interface implementation 0. 5.16 6.10 9.75 6.16 11.
rectangular containment with an arbitrary recirculating flow field.27 and Fig.24 in which the source mesh has the highest steam concentration.25 and Fig. There was initially air in the containment and steam is introduced in the source mesh at the rate of 0.28 show the steam concentration distributions after one time step.2 kg/sec. Figure 5. the steam concentration for the new time step can be calculated in an explicit way. 5. 5.26. 5. The steam concentration distribution at the beginning is given in Fig. 5.149 while the mesh interface implementation gives a physically reasonable solution.26 show the diffusion constant corrections of the mesh point and mesh interface implementations and Fig.25 and Fig.19. Figure 5.23.28 is not a reasonable solution while the mesh interface implementation result in Fig. 5. Since the steam concentration distribution and flow field are given at the start of the calculation. 5. The true solution cannot be reproduced because some of the information about the flow field has already been lost in Fig. The result in Fig. After a while a flow field is set up in the This containment by the input mass and momentum of steam.22 shows the flow field in the containment and more detailed information about the flow field is given in Fig. 5. flow field is used here to test the mesh point and mesh interface implementations of the corrective scheme.3 sec with the corrections in Fig.27 is physically reasonable and even better . 5. 5. respectively. Another test calculation is done for a two dimensional. 5. 0.
5m Ay = 1.150 Ax = 0.0m Fig. 5. Recirculating flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the convective scheme .22.
5955 SOURCE 0.0127 0.151 0.456 0.93 2. 045 0.0348 0.3074 0. 3563 0. 861 1.1229 0.5174 1.7126 0.2629 0. 0735 0.3 095 1.34 0.2460 0.1139 0 .1276 0 .0937 0 1. 5.2216 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 .2474 04 1379 0.207 2.2190 0.0492 0 . 1317 0.531 0.2435 0. 0648 1.4870 0.3957 0. 2214 velocity [m/sec] Ax = 0.2159 0.23.1385 0.5 m Ay = 1 m Fig.3253 0.4920 0 .1422 2148 0.5893 0.
1178 0.1710 0. 5.0620 0.0358 0.1592 0.1750 0.2 kg/sec of steam Fig.0049 0.1859 SOURCE 0.0819 0.152 0. Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .0002 0.1712 0.0031 Ax = 0.0095 0.0298 0.5m by = Im Source: steam concentration in Ikg/m 3 0.0007 0.1266 0.0305 0.1420 0.0346 0.24.0136 0.0011 0.0179 0.0676 0.1085 0.1244 0.
0590 0.1279 0.5m by = Im Diffusion constant in Im2/sec] Fig. Diffusion co. 0396 0.1451 0.1178 0.25.0224 0.0300 0.0600 0. 0.0255 SOURCE 0.3262 0.1701 0.0794 0. 0227 0.. 0086 0.153 0.1694 3106 0.2740 0.0193 0. 5.0 362 0.0937 0.0731 116 0. 0.0 183 Ax = 0.1452 0.0253 0. 0152 0. 0. 2422 0.0440 0.0 278 0.1763 0. 0488 0.1014 258 0.stant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .0311 0.0311 0.1 389 0.1583 0.1854 0.0348 0.C209 0.
0192 0 0.2317 0 0.0 111 0 Lx = 0.1149 0 0 0. 0240 0.0479 0. 0579 0.5m by = lm Diffusion constant in Im2/sec] Fig.9769 0 0 0.0860 0.0 257 0.26.2134 0.0632 0.154 0 0 0.3694 0. 5. Diffusion constant corrections of the mesh interface implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .0287 0. 0706 0.4732 0.2529 0.0534 0.1026 0 SOURCE 0 0 0.0443 0.0518 0 0 0.0215 0.0 76 0.0 923 0 0.
1321 0.1706 0.0448 0.0011 0.2 kg/sec of steam Fig.1847 SOURCE 0.1623 0.1363 0. 5.1243 0.0195 0.1135 0.5m Ly = Im steam concentration in [kg/m 3 ] Source: o.27.1475 0.0046 Lx = 0.155 0.0342 0.0342 0.0107 0.0833 0. Steam concentration distribution after one time step.0016 0.1711 0.0233 0. 0.0735 0.0947 0.0465 0.0003 0.0068 0. with donor cell differencing convection term without any diffusion of correction .1733 0.3 sec.
0010 0. 5. Steam concentration distribution after one time step.156 0.0355 0.0465 0.2 kg/sec of steam Source: 3 Fig.0005 0.1334 0.5m Ly = Im steam concentration in Ikg/m 0.1747 0. with mesh point implementation of the corrective scheme .0029 0.1247 0.1911 0.1493 0.0092 0.28.1797 0.3 sec.1729 0.0937 0.0815 0.1646 0. 0.0042 Lx = 0.1240 0.0229 0.0068 0.0433 0.0287 0.0250 0.0709 0.2086 SOURCE 0.
3 sec. 0.1112 0. 5.0481 0.0815 0.1310 0.0007 0.0228 0.29.0058 0.1243 0.0360 0.0039 ax = 0.1756 0.0003 0.1628 0.5m by = Im steam concentration in [kg/m 3 Source: 0.0947 0.0192 0.0295 0.2 kg/sec of steam Fig.1794 0.0010 0.1487 0. Steam concercration distribution after one time step.1847 SOURCE 0.0722 0.1342 0. with mesh interface implementation of the corrective scheme .157 0.0424 0.1738 0.0100 0.
Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable. =10. 5. The results of the ADI and explicit schemes are compared in the problem geometry of Fig.158 than the donor cell solution.31 shows the profiles of 4 along the dotted line for the time step size of 0. In Fig.3 ADI Solution The ADI scheme is tested to use a time step size larger than the Courant limit. 5. The ADI and explicit schemes give almost identical results during all the transient.5 sec.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. In the meshes at the top of the source mesh. Figure 5. Figure 5. 5. when the Courant limit is 1 sec. 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.30.0. which is exactly the . 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. it has the maximum time step size that can give a physically reasonable solution. 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.30 there is an inflow from the bottom surface at the velocity of v=1.
Problem geometry for comparison of the explicit and ADI solution schemes .~ ~ILI_~_~LI*L~l*IIIU*XI 1 9 IB Source of ' 'A Profiles of 4 = 10.0 c= 1. 5.0 on this line are compared.1I I _~ 1_.YXI~^.0 Ax = Ly = 1.0 Fig.30. v = 1.
0 A iatlai B 10. 0A Explicit ADI 10.0 10.0 10.0 11.0 12.0 Fig. Comparison of the explicit and ADI solutions for along the line AB in Fig.30 the profile of sec and the Courant limit of 1.0 11.0 0.0 10.0A 12.0 11.31.0 11.160 12.0 10.0 coordinate 11.0 11. 5.5 sec 12.0 sec for Lt=0. 5.0 12.5 .01 12.
30 for At=1.0 11.0 .0 12.0 ~ spatial coordinate 12.0 4 sec 11.0 Explicit ADI 10.0 11.0 16 sec 20 sec 12. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig.0 10.0 Fig.0 10.0 10. 12.161 1 sec 12.32. 5.0 11.0 12.0 sec and the Courant limit of 1.0 sec . 5. 11.0 10.
there is some deviation between the two results.32 and Fig. 5.33. Initially. 5. but it damps out rapidly and the ADI scheme gives almost identical results with the explicit scheme.31. 5. 5. . Fig. No crossflow diffusion corrections are involved in the results of Fig. 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.33 shows that the ADI scheme can use a time step size up to five times the Courant limit.
163 12.0 At =0.0 t spatial coordinate B Fig.30 with different time step sizes . Comparison of the ADI solutions for the profile of at t=10.0 sec along the line AB in Fig. 5. 5.5 sec ec 11.33.
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. (2) The models of laminar and turbulent diffusions in Chapter 3 may be adequate for predicting the hydrogen transport in the containment. The . The con tinuity/momentum equations are decoupled from the scalar transport equations and solved by the SMAC scheme to obtain the flow field. The mass diffusion.164 CHAPTER 6 CONCLUSION 6.2 Numerical Schemes (1) There are two numerical diffusion sources. truncation error and crossflow diffusion. in the finite difference donor cell treatment of convection. and turbulence equations are solved separately using that flow field. The thermal equilibrium is assumed and the phase change occurs to maintain 100% relative humidity. or superheated steam as may be appropriate. 6. energy. Crossflow diffusion occurs due to the donor cell treatment of the convection term in a multidimensional problem. The total diffusion con stant is the sum of the laminar and turbulent diffusion constants.
It gives unphysical results for most recirculating flow and coarse mesh problems. ~unx. The skew differencing scheme gives good results for some problems. It is conservative and an explicit scheme can be used without affecting the simple solution structure. Therefore. (2) Two types of approaches. convection dominant. The corrective scheme is based on the fact that the additional crossflow diffusion can be predicted theoretically for every mesh at every time step. . The truncation error diffusion occurs in the flow direction while the crossflow diffusion occurs in the diagonal direction of neighboring mesh points. have been tried in order to eliminate the numerical diffusion. but not always. skew differencing and corrective schemes.165 effective diffusion constants of the truncation error and crossflow diffusion are of the same order of magnitude. the corrective scheme is generally preferred to the skew differencing scheme. The inclusion of corner points complicates the matrix structure and a fully implicit scheme should be used for maintenance of stability. Therefore. The gradient of the scalar quantity under consideration is usually small in the flow direction in comparison with that in the direction normal to the flow. recirculating flow problem is due to the crossflow diffusion. most of the numerical diffusion error in a multidimensional.
i. First. Cx. Implicit (ADI) scheme can be used to increase the time step size and decrease the overall computational effort. are identical in a unidirectional flow. using the characteristics of a quadratic equation.4. The mesh interface implementation is always preferred to the mesh point implementation in a recirculating flow problem.. are tried for the corrective scheme. diffusion and cell Reynolds numbers.e. it has its own limitations due to the following factors.2 imposes a maximum time step size that can give a physically . The graphical method is used to obtain the stability condition in terms of the Courant. dx and Rx . the physical constraint in section 4. (5) The maximum time step size is limited by the Courant The Alternate Direction condition in an explicit scheme.166 (3) Two implementation strategies. The results of numerical experiments are found to be consistent with the Von Neumann analysis. (4) The Von Neumann stability analysis is applied to various finite difference forms of a general conservation equation. The results of various sample problems show that the mesh interface implementation of the corrective scheme always gives a physically reasonable solution with negligible numerical diffusion error. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable. mesh point and They mesh interface.
the ADI scheme is a fractional step method.167 reasonable solution. a steadyThe state solution may never be reached due to the fact that one time step is composed of a few fractional steps. Second. For example. . asymmetry of the given problem may give different results according to the sweeping sequence.
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. skewdifferencing and corrective schemes. etc. distance from the wall). it is necessary to predict the flow regime in order to use appropriate physical models.168 CHAPTER 7 RECOMMENDATIONS FOR FUTURE WORK 7. transitional and laminar flow regimes. time step size. They may still be improved further to get an accurate solution in a transient. The flow regime criterion may be replaced with a model that covers all the turbulent. The model should show proper limiting behaviors as the Reynolds number goes to infinity (turbulent) and zero (laminar). multidimensional recirculating flow problem.g. velocity gradient.2 Numerical Schemes There have been two approaches. A criterion for the flow regime should be developed in such a form that it may be implemented into a computer code. 7. previous time step information. to eliminate the crossflow diffusion. geometry (e. The important parameters may be the velocity..
More tests and validation calculations are required in recirculating flow problems. The conservative form of the skew differencing scheme is complicated and time consuming in comparison with the corrective scheme. The stability of the skewdifferencing scheme is open to question in the explicit. Skew differencing scheme 1. ADI and implicit schemes although only the implicit scheme has been used up to now. It may be possible to develop a reasonably simple form that may or may not include the corner points. The solution in Appendix A may be a useful guide in this work.169 Corrective scheme The corrective scheme has been successful by the mesh interface implementation strategy in the scope of this work. 2. .
34..A.G. T. Engineering Calculation Methods for Turbulent Flow.H. W.A. Briggs." Journal of Comp. Phys. (April 1980). and McDonald. "Hydrogen Distribution in a Containment With a HighVelocity HydrogenSteam Source." Journal of (1968)." Journal of Comp. 372397 (1977).H." Comp. and Eng. Cebeci... Phys. and Whitelaw. "A New Finite Element Formulation for ConvectionDiffusion Problems.A..R. 3. A. C. Briley." MIT Ph. "A Finite Difference Scheme for the Incompressible AdvectionDiffusion Equation..H. D.R. Baliga. W. Transport Phenomena. 94110 Amsden. P. and Harlow. Amsden.. Methods in Applied Mech. H. E.R.. "BEACON/MOD 3: A Computer Program for Thermalhydraulic Analysis of Nuclear Reactor Containments .User's Manual.R. W. 332325 (1970). et al. F. S. 233241 (1975). and Harlow. "The SMAC method: A numerical technique for calculating incompressible flows. F. 24. "Solution of the Multidimensional Compressible NavierStokes Equations by a Generalized Implicit Method. R. Bird. and Lightfoot.N. F. Bloom. Phys.B. J. G." HEDLSA2682 (October 1982).R. "Transport of Turbulence in Numerical Fluid Dynamics.H.E. A. Briley. "A Simplified MAC Technique for Incompressible Fluid Flow Calculations.D. thesis (1979). Bradshaw.. Phys..V. Comp." Numerical Heat Transfer 3. and Harlow. 6. John Wiley & Sons (1960). Broadus. 5473 (1980)." NUREG/CR1148 (EGG208).. A. [2] [3) [4] [5) [6] [7] [8) [9) [10) [11) [12] ." LA4370 (May 1970).R. 6. Academic Press (1981).. Boyle. and Patankar. et al. "On the Structure and Use of Linearized Block Implicit Schemes.. 393409 (1980).. Stewart.. and McDonald.170 REFERENCES [1] Amsden. B." Journal of Comp. D. H. "Transient Effects in Turbulence Modelling.
D... 32. S..F.. 27.. Clark..I.. L. 95107 (1982). Methods in Applied Mech.." MIT Ph..C. P." The Phys.171 [13) Carbajo. Vol." Computers and Fluids.. Vol. G. B. 471484 (1967).H.. "SOLAICE: A Numerical Solution Algorithm for Transient Compressible Fluid Flows. P. Hirt.W. S. De Vahl Davis. C..P. and Romero. Chang. Y. "Coolant Mixing in the LMFBR Outlet Plenum. "VARR2 .I. 1344 (1981). Academic Press (1964)." Journal of Comp. J. "Numerical Methods of Reactor Analysis. thesis (1977).." Nuclear Eng.. Chen. Nielsen. and Eng.S. "Simulation of Containment Cooling with Outside Spray After a Core Meltdown. "Tensor Viscosity Dukowicz.. M. L. Cheng.B. and Lewis. Cloutman. Phys. Cook. 127137 (1982). 4. Cebeci. G. [14] [15) [16] [17] 118] [19] [20) [21] [22] 123] [24] [25) . of Fluids Supplement 2 (1969). "Comparative Analysis of Numerical Methods for the Solution of NavierSTokes Equations. Phys. H.K. "Studies of Numerical Methods for the Plane NavierStokes Equations. and Nakayama." Nuclear Eng. Hirsh. K. "Buoyancy Effects in Mathematical and Numerical Turbulence Modelling.. S. 1 (Nov. and Hansen.D. "Study of Nonhomogeneous Conditions in Reactor Containments Due to Separation Effects." LA6236 (July 1976). T. Chang. 2943 (1976).D. J. thesis (1980).D." Comp. G. Williams.J. "Accuracy of Difference Formulation of NavierStokes Equations. and Ramshaw.L.. thesis (1981). Method for Convection in Numerical Fluid Dynamics. J." MIT Ph.A Computer Program for Calculating Timedependent Turbulent Fluid Flows with Slight Density Variation." Journal of Comp." CRBRPARD0106. 1. "An Evaluation of Upwind and Central Difference Approximations by a Study of Recircul ing Flow. R.. and Design 69. Crowley.. 7179 (1979). "SecondOrder Numerical Advection. 1976). W.H." MIT M.B. and Mallison.D. Keller. Covelli. J. Jr.. M. and Design 69.G. Varadi.S. N.
2. F. "On the Use of Central Difference Scheme for NavierStokes Equations.P. Wiley. [27] [28] [29) [30) [31] [32) [33) [34] [35] [36) [37) Hirt." Journal of Comp." Int.A. and Launder.. C.172 [26] Fromm.E. and Amsden. "Practical Investigation of Convective Difference Approximations of Reduced Dispersion. 528 (March 1979). 339355 (1968). Hinze. 2nd ed.. J. Harlow. and Amsden." The Phys. and Bird. 18." Comp..A. Hirt. Vol. Han... Phys.M. B. Phys. Hamza. Hirschfelder. Gupta. T. F.... Humphrey. 8093 (1968).B. Turbulence. McGrawHill Book Company. New York (1954).. 309312 (March 1975).H. M. "A Numerical Fluid Dynamics Calculation Method for All Flow Speeds. J. "Numerical Calculation of Almost Incompressible Flow. [38) . J. R. R." The Phys.C. Harlow.. "A Comparison of Hybrid and Quadratic Upstream Differencing in High Reynolds Number Elliptic Flows. J.F.O. Gebhart. 197213 (1971). of Fluids Supplement 2 (1969)." Journal of Fluids Eng.H. and Amsden. "Buoyancy Induced Fluid Motions Characteristic of Applications in Technology ..A. "Improved Form of the Low Reynolds Number kc Turbulence Model." MITEL 81020 (June 1981).W. Hoffman." The Phys.. 3. and Golay. C. 8. 17. Methods in Eng. 29....H. 1952 (1975). A.A.W. 8195 (1981). 557573 (1980). A. Molecular Theory of Gases and Liquids. "Computer Studies of Timedependent Turbulent Flows....W." Journal of Comp. Harlow. M. Phys." Journal of Comp.. "Behavior of Buoyant Moist Plumes in Turbulent Atmospheres. Vol. Phys. J. G. B. C.. 15. (1975). Vol. 101. Methods in Applied Mech. for Num. A.The 1978 Freeman Scholar Lecture.." Journal of Comp. and Eng.. of Fluids. "Heuristic Stability Theory for Finite Difference Equations.O.H. F. of Fluids Supplement 2 (1969). "Numerical Calculation of Multiphase Fluid Flow.E. R. and Marohar. Curtiss.
P. 103. and Launder. Frankfurt... Jones. and Golay.." Int.E. D. 301314 (1972).. 293312 (1980). 15. Reece... Vol. "The Calculation of Low Reynolds Number Phenomena with a Twoequation Model of Turbulence. 456460 (Sept.E. and Eng. "Analytical Modelling of Hydrogen Transport In Reactor Containments. M. and De Braemaecker." MIT Sc.H.H.. Heat Mass Transfer. "A Modified Form of the kc Model for Predicting Wall Turbulence. Langer.J.P." Int. B. 148) [49) .P.. 269289 (1974).K. Jr.E.P. and Spalding." Trans. John Wiley & Sons (1978). "ADI Solution of Free Convection in a Variable Viscosity Fluid. part 3. Vol.3633. M.. Lam. V. Jones. G. Vol. Cl..A. 16.E. "Experimental Investigation of Hydrogen Distribution in a Containment of a Light Water Reactor Following a Coolant Loss Accident. Methods in Applied Mech. J. W. B. 16." Comp. of the ASME..W.." BF63. Battelle Institute e. J. 537566 (1975). 68. Steam Tables. and Rodi. Heat Mass Transfer. J. V. K.." Report submitted to 2nd International Topical Meeting on Nuclear Reactor Thermalhydraulics (March 1982)." Comp. G. thesis (1983). 3.B. Keenan. and Eng. translated (October 1980).. Launder. "The Numerical Comparison of Turbulent Flows. "Practical Evaluation of Three Finite Difference Schemes for the Computation of SteadyState Recirculating Flows. Manno. and Keyes. C. Phys. W. M. et al. and Bremhorst. and Launder. F. Methods in Applied Mech.D. J. 11191130 (1973). Launder. [40) [41) [42) [43] [44) [45] [46] [47] Leschziner.G.." J. B.. 221239 (1974)." Journal of Comp. 1981). Vol.173 [39) Houston.V. 23. B. "The Prediction of Laminarization with a Twoequation Model of Turbulence. "Progress in the Development of a Reynolds Stress Turbulence Closure. Manno. "Analytical Modelling of Hydrogen Transport in Reactor Containments. W..G. Fluid Mech.
Hemisphere Publishing Corporation (1980). Methods in Applied Mech. Methods in Applied Mech. of Fluids Supplement 2 (1969).K. "Convective Difference Schemes. Computational Methods in Engineering and Science. P.. Raithby.." The Phys.. 1975).D." Int. "SkewUpstream Differencing Schemes for Problems Involving Fluid Flow.. Patankar. 541550 (1972). Phys. 386389 (Sept. and Spalding. and Eng.B. 9." Journal of Comp. Roache. 191206 (1974).K. [52) [53) [54) [55) [56) [57) [58) [59] [60) [61) [62] [63) . J. for Num. 4. 2. Turbulent Buoyant Jets and Plumes. T. Albuquerque. and Sherwood.. McGrawHill. Inc. Methods in Eng.. Pergamon Press (1982)." Trans. and Torrance. (1977). Vol.J." Computers and Fluids. Mass and Momentum Transfer in Three Dimensional Parabolic Flows. Reid.. Rodi.C.174 [50] [51) Nakamura. Numerical Heat Transfer and Fluid Flow. D. S. Heat Mass Transfer.D. K.. 15.. N. 10. "Upstreamweighted Differencing Schemes and Their Application to Elliptic Problems Involving Fluid Flow. "On Artificial Viscosity. G. "Numerical Experiments with a Difference Model for the NavierStokes Equations (Turbulence Model). Roberts. W. "A Note on the Empirical Constant in the KolmogorovPrandtl EddyViscosity Expression.D. P. New Mexico (1976). S." Int.. "Convergence and Accuracy of Three Finite Difference Schemes for a Two Dimensional Conduction and Convection Problem." Comp... W. Roache. 169184 (1972)." Mathematics of Computation..B... Computational Fluid Dynamics. R. 153164 (1976). 20 (1966). Raithby. 9.J. Hermosa Publishers. of the ASME.V. "A Calculational Procedure for Heat. Raithby. Runchal. New York (1958).E.. The Properties of Gases and Liquids.. G. 75103 (1976). S. G. Rodi.O. J. W. "A Critical Evaluation of Upstream Differencing Applied to Problems Involving Fluid Flow." Comp.V. and Weiss. Schonauer. John Wiley' & Sons. A. and'Eng. Patankar.. K. 17871806 (1972).
S. 411428 (1980). "The Numerical Treatment of Advection: A Performance Comparison of Current Methods.. Spalding.B. No. 3. Vol. For Num. "Comparison of Finite Difference Journal of Computations of Natural Convection. C. and Valentin. No. "Proposal for a New Discrete Method Based on an Assessment of Discretization Errors.. J.. K." Journal of Comp.A.H.Dec. K. Y.B. "Diffusional Properties of Multicomponent Gases.E. Trent.. 2..B. 20.D. 46. R. Wilke. Stubley. D. (Jan. 3..G. 1981). A.." Chemical Engineering Progress. "A Novel Finite Difference Formulation for Differential Expressions Involving Both First and Second Derivatives. M. Vol.. J.. Phys. A First Course in Tur [65] [66) [67] [68) [69) [70] [71] Torrance. "Simulation of Convection and Diffusion Processes by Standard Finite Difference Schemes and by Influence Schemes. "TEMPEST: A Threedimensional Timedependent Computer Program [72] for Hydrothermal Analysis. 4." Int." Numerical Heat Transfer. Stubley. Raithby.. and Eng." of the National Bureau of Science .. and Strong. Smith." Comp. . 95102 (1950). H.L. MIT Press (1970). 551559 (1972). D. A. Methods in Applied Mech.L. 411428 (1980). G. A. Methods in Eng.. and Eyler.W. L. bulence. and Hutton. 35. and Lumley.175 [64] Shin. 1968).D." FATE80114 [73] (Jan.D. Tennekes." SAI74509LJ.. "Numerical Analysis of Fluid Hammer Waves by the Method of Characteristics. and Woolner. MatheResearch matical Sciences.. "Development and Validation of a TwoVariable Turbulence Model.A. 4 (Oct. 220237 (1976). R..M. J.B. Strong. G. 72B. Stuhmiller.. G.. Budden." Numerical Heat Transfer. 1974).J.D. 153168 (1982).R. Raithby. G.
c. a= k PCp C =2 u 2A.1 convection occurs in the x direction and diffusion occurs in both x and y directions. pc u aT = k(a2T + 2aT 2 x2 p x ax ay (A.1. The boundary conditions are specified in Fig.2 is simplified further by substitution of 4 for T as follows.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.3) ax . Therefore.3(B) and the constants can be grouped into one constant.2) Equation A. Eq. as follows. 5. A.176 APPENDIX A ANALYTICAL SOLUTION FOR THE PROBLEM IN FIGURE 5.a2T aT 2 + a2 2 2c ax ax 2 y ay (A. A..1) In Eq. 2 x 2 a2 ay = c 2 (A.
A. A.3 is reduced to the following form. A. I d2X + 1 d 2Y 1 = c2 C (A. c 2 X cx 2 = (A. which are functions of x and y only respec (x. ables. The variable 1 is separated into two vari X and Y.7) d Y c .4) The boundary conditions should also be expressed in terms of 4 as given in Fig. Equation A. tively.6) dx dy The two terms on the left hand side of Eq.3 can be solved by the method of separation of variables.8) dy where a > c > 0 .y) = X(x)Y(y) (A.a =  (A.5) Then Eq.177 where 4 = eCXT (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.1.
3(B) where 0 is given by.0 0 = 0 x Fig.= YO T 0 ii ii i h YO + co = 0 4 = 0.178 . Boundary conditions in terms of 4 for the problem in Fig. 4 = ec T . A.1. 5.
16) . A.14) (A. 13. dY dy y=O Y=O 0 (A. A.12 and Eq.7 and Eq. the constants in Eq. The boundary con dition at x=0 will be considered later. C3 = 0 (A.11.12) (d dx _cX) = 0 x=xO (A. A.9 and Eq. X = Cleax + C2 eax (A.9) Y = C3sin8y + C4cosSy (A. A. A.15) ny 0 = n C2 _ n+c 2 e nX 1 n (a n 9 C) (A. A. A.179 The solutions to Eq.10 should satisfy the following relations. 13) In order to satisfy the boundary conditions given in Eq.8 can be easily obtained as follows.1 may be given in terms of the variables X and Y as follows.10) The boundary conditions in Fig. Eq.11) dY dy=Y 0 = 0 (A.
Th 0 2. . Therefore.3 . .y) = e cx (x.18) n n Since the boundary condition at x=0 should also be satisfied.) (A.2.y) where nr (A.1 are determined as follows. the constants A. 17) T(x.0 2 A n Th h nr 1 n 1+( a c )e n 1 a +c 2a x . the final solution can be expressed as follows. 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=1.19) .180 where 2 =c 2 + n (O) n 2 yO =c 2 + 8 2 .
1Eq.1 and B. When the tilde phase of the SMAC scheme and other scalar transport equations are solved by the ADI scheme. The time step size limitation in the ADI scheme comes from the semiimplicit treatment of the nonlinear convection term. there are six sweeps in a three dimensional problem and four sweeps in a two dimensional problem.iillaY~i~~l_ ^YII^^~II ( . They are formulated for a natural convection problem where the gravity acts in the zdirection. 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.i. B. Therefore. Since there are two steps involved in solving the momentum equation in each direction. the computational effort for one time step is greater than that of the explicit scheme. B. The most important sweep should be given the first priority and most updated information. The sequence of the sweeps is arbitrary and depends on the problem under consideration. The ADI scheme is applied to a two dimeraional momentum equation in the following Eq.4.3 are the z and x direction sweeps of the x direction momentum . Equations B. 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.
2ui+.2ii+ Ar 2 + Wilj+ .2) n+l i+j * . 2 Az + Ui+ 1 P  n  i+j i+ jui+4j (B.2 and B.u n+l ui+2 2 n+l Ar2 +2u j n+l ui j . ar .ui+ j At  n+l * u i+lj i+ .1) * * n * *  ij+ At Wi+1 j+ ij+ + i+Ij+wij+ uij+Wilj+ 1 P n~ .182 equation. u n * u j i+ Ui+D Dt j + * n i+ j+ Ui+ j .Rxi+ jui+ jui+ j RX * n+l (B.Wi+ Az j Ui+ j 1 ui+ j+l * . sweep of the z direction momentum equation is treated as the last one in order to be given the most updated information.4 are the x and z direction The z direction sweeps of the zdirection momentum equation.3) .P z  n * RZij+ ij+ ij+ (B. Equations B.u Lr * n+1 .
4) Equations B.3.2. and B.183 n+l i+ * wij+ * i n+l +lWij+ * n+l zW. B.+ n+1 n+l Az 2wi + w _P z P0 zP g z RZ n+l ij + wij+ 3wij+ (B.1. . B.4 have given almost identical results with the explicit scheme for a time step size below the Courant limit. w At n+l Wi.
All the inflow velocities The component velocities are used to calculate the crossflow diffusion constants for eight subspaces in a three dimensional coordinate space. 4. There are When there are inflow and outflow.3(A).2. two different cases treated separately. In the second part of the program the flow split is considered such that every outflow velocity is partitioned into four components in a three dimensional case in proportion to the velocities of the four neighboring surfaces in contact with the surface under consideration.3(B) and Fig. y and z directions.4.184 APPENDIX C COMPUTER PROGRAMS The files.8. the calculational logic of the mesh interface implementation of the corrective scheme in section 4. or all inflows in x. When there are two outflows in any of the x. 5. is the program for testing Fig. Then four crossflow diffusion constants at every interface are summed up to give a total crossflow diffusion constant . 5. the crossflow diffusion constants can be calculated in a simple way as in the first part of the program.. 'progl fortran'. Fig. y and z direction mesh interfaces. 5.8.3. are the computer programs for calculating the numerical values of the analytical solution in the problem geometries. a flow split occurs as shown in Fig. The file. 'difprg fortran'. 'prog2 fortran' and 'prog3 fortran'. are assumed to be zero.
has four neighboring subspaces. 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 ADI scheme for the momentum equation is given in Appendix B. the positive x direction.185 because one axis direction. 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.
2.NY  0.I=l.XO)/(ALNC) GO TO .4ALNX) ATOT .NX.O) ALNX .0 44 YY I IIPI*(UJ0.0 IHPI DO SUM 10 JJI.215.0*TII/PI PI/2.COF*DSIN(II*HPI)/(1.3) SlOP END PRO00210 PRO00220 PROO0230 P1ZOOO240 P1I000250 PROOO260 PROOO270 PROOO200 PROOO290 P11000300 PROO0310 PRO00320 PRO00330 PROOO00340 PRO00350 PROOO360 PROOO370 PROOO300 PROOO390 PROOO400 .OALNOEXP(ALN.EO.14 55 ATOT * THIODEXP(ALN*XO)/2.3F7.000190 PI00O200 NY*OX YSO v PI*PI/(YO. AAN*2.OZ) T(tO) DIMENSION REAO(10.N FILE: PROG FORrRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT REAL*0 (AH.NY.26 II IF  IP  I GO TO 55 ALN = DSORT(CSO*II*II*YSO) (II.1 ) DX.O*ALNXO)/(ALNC) COF * CCOFF/lI AAN .NY) 100 FOnMAT(IOX.ALP II FOIRMAT(F7. C*C XO " NXOX YO * PROOOO0 PROOOO20 PROOO30 PROO0040 PRO000OOOO P14000070 PRO000OGO P11000090 POOOOT100 PROOOO10 PIIOO0120 PRO00130 PROOO140 PROOO150 PRO00160 PROO000170 PRO00100 P1.U.Tl.IALNIC)*DEXP(2.100) (T(I).1415926536 C = U/(2.2) PI a 3.YO) CCOrrF 2.5)/NY ADD * ATOTTDCOS(YY) SUM T(d) v SUM + * ADO * DEXP(C*XO) 20 CONTINUE SUM 10 CONTINUE WRITE(G.5F11.04ALP) CSO .0 00 20 IPsi.
OTHIH/PI 11PI = PI/2.5+YO 00 20 IPs1.0 SUMO SUMI r 0.X = 1.0Z) DIMENSION TII(10).3F7 .0 NY2 XXX NY/2.0 = YO/2.2.+ALNX) BEN 11*PI/YO YOO .0 .DCOS(BEN*YI) YI Y22 • DCOS(BENY2) OCOS(BEN*Y3) Y3C Y3S .2) PI .XN = XN .0 X = (JJO. TL(10).0 SUM3X 0. IC1l REALi0 (AII.5)*DX = YC = (JJ.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN .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 = 2.0 SUMaX SUM4 XN vI Y2 Y3 Y4 • 0.FO.0 = 0.5) XALO = DEXP(ALN*XO*O.1.21 II = IP  I ALN = DSORT(CSO + IIII*YSQ) EXALO = DEXP(ALN*XOO.(ALN#C)*DEXP(2. U.5*YO . GL(10).215.3.11) DX. 11 FORMAT(F7.U/(2.NY2 0.DSIN(BEN+Y3) Y4C .21)*DX = X 4 YO/2.O) GO TO 55 ALNX .0 SjU3 " = 0.OCOS(OEN*Y4) Y4S DOSIN(3EN'Y4) ADO * AAN*(EXALO + ALNX*XALO)*YOO PROOO520 PR000530 P1OOO00540 PR000550 . NY.0 SUM2 = 0.0  RTO DO =  OSORT(XXX) 10 JJ=1.DCOS(BEN*YC) .p A FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 MPI.0 .OALP) CSO = C'C XO * PROOOO O PRO00OO20 P R000030 PROOOO40 PROO00SO POOG0050 PROOOO000 PROOOOOO PROO0090 PIR000 100 P11000110 PRO00 120 PROOO 1. NX. ALP READ(10. 1415926536 C .0 X + XO/2.5) EXAL = OEXP(ALNOX) XAL = DEXP(ALN'X) EXALN D EXP(ALN*XN) XALN = DEXP(ALNXN) IF(II.COF*DSIN(II*HIPI)/(I. TC(IO) TH14. GH(10).O.
NY2) WRITE(.EXALN + ALNX*(CALN)*XALN)*Y3C EN*AAN#(EXALN .5) SUMI*DEXP(C*X) SUM2*DEXP(C'X) (SUM3X+SUM3Y)*DEXP(C*XN)/RTO (SUM4X+SUM4Y).100) (GH(I).NY2) = t.00) (GL(I).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 .NY2) WRITE(G6. 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.NY2) = I.SFII.II.0 ADOvY * 0.5 A002 * TMIIIXAL0.(" % FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 ADOI w AAN*(EXAL+ALNX*XAL)*YII AD02 . 1OO) (TH(I).0 ADO3Y * 0.3) STOP ENO SUMO*DEXP(C*XO0O.NY2) WRITE(6.I WRITE(6.tOO) (TC(I).AAN*(EXALALNX*XAL)*Y22 ADO3X v AAN*((C+ALN).I 1OO FORMAT(IOX.0 AOOX a 0.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.II.5 ADO3X 0.tOO) (TL(l).5 A)DO * TIHl'XALO*O.
G1(10).0*ALP) CSOQ C*C XO = DX'NX YO = DX*NY YSO = PI*PI/(YO*YO) CCOFF = 2.DCOS(UENOYC) YIl DCOS(OEN'YI) Y22 .215.0 XC=Y0/(2.0*RT3))/NY2 Y3=(JJO.XO/RT3.0*RT3#NY2) Y2=YO/2.3F7.0)/NY2 X4XOYO/(2.2) PI = 3.O' XO/RT3)/NY2 XI=(dJ0. ( 10).OtTIIIM/PI HIPI = PI/2.5)*YO/(2.0 SUM4Y = 0.0*RT3)+(J0.0 .5)*(XO/RT3YO/6.0 SUMI = 0.NY2 SUMO = 0.0 NY2 nRo P000O070 13100000 PROOOOOO PROOO0000 PRO00110 = NY/2.2. TL(IO).1415926536 C = U/(2.02) 111(10). TC(1tO) PROO0020 PROO0030 PROO0040 PnOOO050 PROOG000 READ( 10.0.O*RT3)+(JJO.SO5T(2.5)*YO/(2.0)/NY2 X2=(JJ0.0 SUM4X = 0.o) ALNX = (ALNwC)tDEXP(2.S)*(XO/RT3YO/6.0) PROO000120 PROO000130 PROO0140 PROOO150 PR000O160 RT3 = C SORT(3.0) DO 10 00=1.5) * (YO/3.0 PR00O170 PR000100 PROO0190 PR000200 PlO00210 SUM2 = 0.5)YO/(2 .0+(dJO.S)#(XOYO/(2. TIIl. ALP PROOOO10 GL( O). I 1) DX.0 SUM3Y = 0. NY.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN = COF*OSIN(II'HPI)/(I. rORMAT(F7.0*RTJ))/NY2 YIYO/2.NX*DEXP(ALNXI))*YI I ADD2AAN*(DEXP(AI. FILE: PROG3 FORIRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT DIMENSION C 11 REALtO (AH.DCOS(iEN*tY2) YV3C = DCOS(BEN'Y3) Y3S = DSIN(BENtY3) YAC = OCOS(3EN*Y.O*NY2) .EO.0 SUM3X = 0.0*NY2) X3YO/(2.+ALNX) OEN = II'PI/YO YOO . U.5)*(XOYO/1r3)/NY2 YC=(JO.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 .C.*0RT3#NY2) Y4=YO/3.RT3)+(dJO.0(JJ0. NX.5)YO/(2.0.5)*(XOYO/(2.(dJ0.4) Y4S = DSIN(6ENlY4) ADDO=AAN*(DEXP(ALN*XC)4ALNX*DEXP(ALN*XC))*YOO ADDIAAN9(DEXP(ALN*XI)+AI.22 II IP I PR00360 PRO00370 PROOO3UO P(1000390 P000400 ALN = DSORT(CSO + II*laIYSO) GO 10 55 Ir(II.C P11000220 PROO0230 PROO0240 P1000250 PROOO2GO PROO0270 P11000200 PROO290 PR000300 PROO03 10 P 11000310 PROOO320 00 PRO00330 PR00OO340 PROO0350 00 20 IP=1.
NY2) tOo rORMAT(IOX.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.0 ADO3Y = 0. = 0.3) STOP ENO .5+SUM3Y*0.0 ADDOX a 0.11.I= . PROOO6tIO .NY2) WRITE(6..I=I.5)*DEXP(C*X4) 10 CONTINUE WRITE(6.0OEXP(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.1OO) (TL(I).5 ADO3X PROOOSGO PROO00570 PR1000500 PROO000590 PROOOGOO P1100000.I1.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) (GL(I).5*RT3)*DEXP(COX3) GHi(JJ)(SUM4X#O.I=.0 77 SUM1 . 100) (GI(I).NY2) WRITE(G.NY2) WRITE(6.X4) ALNX*(CALN)*OEXP(ALN*X4))*Y4C ADO4Y.5 ADD2TtII#DEXP(ALN*X2)*O.100) (TC(I).5 ADDH11H*DEXP(ALN*XI)*0.NY2) WRITE(6.5*RT3+SUM4YtO.0 ADO4Y = 0.100) (TH(I).SF11.OEN*AAN(DEXP(ALN*X4)4ALNX*)DEXP(ALN*X4))*Y4S GO TO 77 55 ADDOTIIII*DEXP( ALN*XC)*O.
5) DELR.GT.0 RFGTI .O) IF(WP.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 .WM.UU/DFLR PPZ * VV/OELZ PPT .OR.VELZ(0).O.0 IF(UM.VP.VELT(O).LT.L1.OR.0 OMT .DEL'TII.ISt + IS2 ISO IS3 + IS4 ISC ISS5 + ISG C UI = 0.t = 1 = i v 1 * .( WM W2 0.O.DIFRA(0)1.WW/DELTII PSUM .2.VM.UU*DELR*( PPZ + PPT )/PSUM DIFZ * VV*DELZ*( PPX + PPT )/PSUM DIFT .DIFZ'IS3 * DIFZ*IS4 RFGr GO TO 00 AOS(UM) + ABS(UP) DI F00170 DIFOO IGO DIFOO00190 DIFOOlOO ISI IS2 IS3 IS4 155 ISG .EO.5.ISC.EO.DELZ.O.5+( UM U2 0.PPX $ PPZ + PPT GO TO 500 IF( PSUM.0 ) C DIFR .4 ~ FILE: O FI'IlG FOITIAN A VM/SP COINVERSATIONAL MONITOR SYSTEM PACE 001 DIMENSION VELX(0).GT.5*( VM V2 0.2) UU * UI + U2 VV .EQ.UP.O 151 4 0 ISS1 O 156 O DIFOOOIO DIFOOO20 DIFOOO30 OIF0040 011:O0050 DIFOOOGO DIF00070 DIFOOOTO DIF00110 RFGRI = 0.0 RFOT 2 0.O.IS .0) IF(VP.5*( VP WI * 0.) ISA .0. bt ts .DIrn*ISI .WW4DELTII#( PPX + PPZ )/PSUM C RiFGII .0) IF(VM.I 01o00200 DIF00210 DIF00220 DIFOO230 DIFOO240 DIF00250 DIF00260 DIFO0270 DIF00200 OlFOO290 DIFOO300 DI F003 10 I'I .O.EO.VI + V2 WW = WI 4+ W2 PPX .5*( WP C IF( ISA.GT.0 n I FOO 130 OMTI m 0.LT.0 RFGCR 0.0) IF(UP.0) IF(WM.2.O 15S3 .0.IFZA(0).0.WP 1SI * 0 152 .5*( UP VI 0.DIFII'IS2 rGR RfGTI .UM.DIFTA(0) REA( 10O.O.
0PET) Vt .FILE: DIFPRC'.0.0PFz) W2 &1EX'&( 1.V(LT(I).0 ) GO TO 600 DIFRA(I) .PFX*PFT V2* I'F X'PF T V2.0.0.VELX( I)/I)ELR PPZ . oi rooG to Dl F00G20 D1F00630 01F00640 Ut 'PFZ .0.VELZ(I)/DELZ PPT .0 oiETA(I) . Z(2) vrI.0PFZ) DIFO00650 oDI roofGo oir00670 nir006s0 mrof0069 ot rooioo DIF00710o VEIX(S) VELX(6) VIELX(7) VELX(0) VEL7( I) V El.0PFT) U I PFZ (I 1.O..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.0PFX)#(t.( 1.0.oPF7Z)'PFT UIs .*( PPX + PPZ )/PSUM GO TO 30 600 OIFRA(I) .( .0.0PET) Vi*PFX*( 1.) U2 4PF Z PET 1J2'( 1.0PFX)*PrET V2*( 1.0 Pr T Ir (UI+U2).0PFZ) W20( 1.0PFT.0PFX)'PFZ Wt'(I.09'FX)#(t1.0PET) Vl*( 1.NE.0PFT) V t.0 (Vl*V2).0PFX)o( t.0OPF Z) W2 *P FX' IF Z WI'.IPFZ)s( 1.0PFX)#PFT V2( 1.0PFZ)'$( .DELTH.PrT )/PSUM DIFTA(I) .X(I)'DELR*( PPZ + PPT )0 'PSLJM DIFZA(1) m VELZ(I)*DELz#( rrX +.0 oirZA(l) .0PFX)*PFZ DO 30 11.VELT(I)/OELTH PSUM a PPX + PVZ + PPT IE( PSUM.NIE.OPFX)O( I.o 'FX'( 1.PFT u i.0 30 CONTINUE RFGRI RFGR RFCTI  DIFO 10 tO 011:'01020 DIFO 1030 01 FOt040 DO 1050s DI FOtOGO DIFO 1070 01000 DIFO 01 FOtO0J0 OIFOi tOO DIFRA(5) v DIFfIA(t) aDIFZA(2) + DIFRA(6) + DIFRA(2) + OIFZA(3) DIFPA(7) + OIFRA(3) + DIFZA(6) + + + OIFRA(fl) OlfRA(4) + DIFZA(7) r % I .VEU.0 PPX .0 Ir (W1#W2).EQ.( i.0 0.0PFZ)*PFT U20( 1.PFT) WI#PI'X#PFZ WI'*ix( . ronTRAN A FILE: CON4VERSATIONAL MONITOR SYSTEM DFPRG AVM/SI' FOTRAN DI F00560 DIF00570 PACE 002 PG 0 OMT GO a 0IFT'156 TO 999 00 PFX PF Z *0.NC.Z (3 ) VELZ(4) V EL ( 5 ) 011'00720 DIF00730o DI F00740 ui roo75o DI F007GO ui FO0llO DIF 00700 0D F00190 0117OO8OO OIFOOII10 o)1roos~o OI F001130 DI FOO 40 otrooaso 01 roooGo Dl FOOD 70 VELZ( 6) VELZ(7) V EL Z( 0) VrELT( I) VELT(2) VELT( 3) VELT(4) VELT(5) VELT (6) VELT (7) VELT (0) Dl FOODO DIFOOR90 0 fF00900 01 FO009tiO 01F00920 DIFOO930 oDIF 00940 OlE 00950 Di FOO960 Dl F00970 0 F0090 01 F00990 oiro0tooo W2$( t.0PFT) U2*PFZ*( 1.
2) 7 FORMAT( ItII .OMT I.6F5.RrGT I.OMT GO TO 200 500 WRITE(G.2X.20X.511S ORRY) 200 STOP ENO DIFOi IfO DIFOI 120 DIFOI 130 oDIFO140 DIFO 150 DIFOI 160 DIFO 170 DIOt t0100oo oIro 190 DIFO1200 DIF01220 I I .2) G FOIMAT( III .RFGT. 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 .I.RFGR.7) 5 FORMAT(9F5.
I KN a (X(I() X(K) RE TURN END  X(K41)*COK))/O(K) VAR09Z350 VAflo9:160 VA1109370 VA11093001 VAI.XN) DIMENSION A(N) .0t4O01 X( I ) I~) 13( ) 0(i1) NM I *N I 00) 2 1I.o(N).O(N).1O9:190 VA1109"IO() VAn0094 t VA11O0d1O '4A110943?0 VAIIW14 40 VAfl0O450 VAI*0946) VAI?94 70) VAR0O9400 VAR094 90 VAIR095C VAflfO!'j 10 VA110IJ520 .0.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.NMI .D.P(400).C .SUarIouT INE TDM(A .C(N).
I vtR I5S0 rooOT 00G2 UC 'S(U(]K) ) VLRi 1210 vAR1E230 VAR E20 U(3KXX) ) LUGc .1) GO T0 49998 SOLUTiON FOR SIE.JSP2 DO O40000 DO 4000 K.KBP2 " (3.P.FFX ( .LE0.FX(3.LT.NPC o1 ]F(K..UG.0.:X GO TO 40000 "0OO D.K) z 0.L.2D LUGRDX/(AUGOX +~DG'RD)' .EO.LT. DA'' G..K2NC JKZ2 . 1) GO TO 400 IF(K.L!S( (]KZZ) ) .P2) GO TO 40002 (U(iK)*U(KX))) LUG = LS(O.O.E1O 0005 FP" * IF(.S(0.G  V RIS150 VAR. DNF.E0..0 D'rF2( I .5(V( K)*:(.PM.3. iB10 .E0.xx'.K) * 0..CR.. C C C C U AND v LDE VELOCITY EDU.S( 4 .S . I O50 VI8090 K1 K12 2 KEP CONSTA'NT CLLCULATION C CRO5SFLD'.C "hT.) G0 TOrOOiC D'.VGRD2/(.EO.]K .E230 .ItD.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.F2(J.0 OOC COI~'T2]'UE C VAR P182 vLR a210 VAR E220 VAR.K t KXX a jK . D2).K) a LUGDX.I.G.)K2NC " (Ki)'NWPC .C . 1CONV..KS 2.KZ2))) .UG.K) • AVG.TIONS SECT ION AD !ST ]F(ICLL].I.DFFUSI'O =.
a uJD(2K) )0.1. UD( 3K) UL UR UT Uo PK) UD( 3KP) U0( 3KMS) DCR DCL D!FFCD(IKL) D) F CO( KL.S(UftH)(UCUQ)  VL21S660.'O('.5 UODUMKS) 16D(1PK) * .I DIFFCD(3KL2) D!~FFCD(IKL3) *0(!. . 054 (K.KSPI I KL I lK z lK 1A 0.R2CLI...2)GO a TO 29990VALR iS.I) K2WC' + II)K NC *K2NC "00Z0100 K=2.E003 DO E0100O ) K GO TO50205 222.C RD20 3KS0 S3G'~ to VAR 8340: VARIS5360' RDZDRD2 VAR i !:0.ET/TADD3 VAR IS420 VLO IS440 3F~lDw. VLRIE6t0 V4RI15650 VAR 15690 ULWM'(uLLIC) ( IR'(L'C4UV) VARIE *7 DO.0.5 PC) ').]BR ( '. ( '(K ULH C CON'.1 SRKSrI a * N BUOJY a LP0E7.PKKNUO(IPK) ). KS I KLS a3)K I KP a IK  NP 2 I N1PC VAR 185501 YLR 15560 VAR 1C570 VARias590 VAR 86201 VARi1Es30..5 )r0.0.I ) I KIPC # (KI)'tN'PCL VAR16D VAR 11510 VARi18520 !PK a 31. 3F(3~.E.P( TJ"'.
o2m.% At t x NdE ONCA + lA b + A t + OZ~N(Lt) OdAN 00gla aVA 06!.)10 6666a 01 0 MfN i NGO 07 6 "VA ZNOO NCO ZOM( + (Xt)0M £ ) O~t6tVA (~ ~ .+fl)IdM Y(Qd..(?nOn) i13( fl2f) unc.'N(tX) + 06Md~7A 1 Et a 6Nr NZ At 2 SW SWA r r ).(iim1 (..OSZ6 M7A OdI'N Al dN O7I~t 17A OdM ( A) ONZ Nt + 1IOZN4(Vt) * + 3NZ)(11) + t bi > >4 E A c TA . (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'.Cw'))'A.t) 2 td9N'Zt oo0to oa 3flN tLNOD s0?_O3 M 6 dVA + NAcns + ZntO * (Nt)d(dXt)d ). 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 )*.(.=OStd7A 0506LdV  ~ ~ ~ ~ Hm )CC7r~! Sall*~ ZN07J ±..( (nf1l)(Himlss7 (In3n)(uLM)S9V ) ZNQ.) .8~ VA 0899tMVA (oaQN 3NC647.uQ + + t (Nt)0ftl x = XIt)f ( oo.vt Z1t0 7 Z(J~Si0(o(nsn)(a)a (1n.onH.7A 0060tMIA 0L6Sgv7A 0968LVA I 3a CA~ SimI) 0L96 t'7A O9S63tM7A j0 0 VA d061Nt) AoflsaNx0As + Od.o.(onin) M0'a(On.
K2NC .ROT * 0.C1 .]K .5RDZ*( VRT * ASS(wRT) .IBR DO 81100 K2.5RDX( URT 1 + SIGMARDXD.NWPC DCT 3 DIFFCO(IKL+1) DC.VRE + ABS(WRB) 1 * SIGMA*RDZD.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 .C1.KER D00 82000 12.AeS(WRB)) .( DCT + DCS ) C1(J) a 0.IKBR) CALL TDM(A 1.W(1K)RDT 82000 CONTINUE C CALL TDM(A .KBR + K 81100 U(IK) a X(IX) C DO 82000 K=2.5*RD2( WRT .16PI 1K a 1 * (]i)K2NC + (K1)NWPC IX x (K2)ISR + I 1 E2100 W(]K) * x(Ix) 00 823000 K2.SIGMARDXD*DCL ACBS(URT) .KSPI ]K = 1 + (11)K2NC + (KI)NWPC 1X a (12).X.5 AI(J) a 0. C 00 l8100 I=2.( URT EAS(URT) ) .5.X.5( U()]KS) * U(IMKS+NWPC) ) Ai(J) = 0.DI.RDX.ABS(VRT) ) .DIFFCO(IKL) DCL a D0FFCD(IKL+2) URT x 0.5*RDX( ULT .KBR DO 62100 1=2.5( U(IK) + U(]KP) ) ULT C a 0.8e1.]K + K2NC IKP a 1MKS 1K + NWPC ]K K K2NC NWPC 1KIAS = C DCR .IMKS IKt.( DCL * DCR ) C1(J) z 0..1 ]K £ + (1])K2NC + (Ki)NWPC ) VAR 19260 VAR 19270 VAR19 280 vLRlg2O VAR 19290 VAR19300.SIGIArRDZD3DCT 01(J) a U(IK)ROT 81000 CONTINUE C .ULT * ABS(ULT) 81(J) .$ ]K .KBR) C DO 82100 Ka2.1SP J a (K2)]IeR * 1 .0.SIGt.SIGMhARDZDODCB E1(J) = RDT * 0. VAR19310 VAR19320 VAR19330 VLR193AO VAR19350 VAR19360 VAR1S370 VARie380 IKL a C 1 + (1)'K2NCL + (K1)NWPCL IPK .'5 WRS (W(1KMS) * W(IPKNWPC))O.5RD02( WRe . = DIFFCD(IKL+3) WRT 3 (W(3K) + W(iPK))O.'RDXDDCR D01(J) .ABS(ULT) ) .
.LG!1CE)PA HIM.0xCd. )'ZQdS50 )Zc!S*  a ' a  (HS.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.' ).3nwNtJ.v4)tS ~m * io1.L 0.Htin 0~Q0~*/'.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 ).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( .'%)SS'7 + HSA . z (r 01)~S~7~).)ssv .fl.v%4!)iS +.~) Al NP 1=I000ca 00 .(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.Q ( (S'feXtiM 094=J7A OSZO0t7A (>I).(HL')S.40 C0~7A O CO0d7A NAoC1s (A(I)a .mNtAz + + ONA(lt.~tSC(Hifl)ss b3(3 ). ). 3)3QO.ZC(3aE1~S (d~t)d  00oo' ((HIiA)SS7 I )'ZQZ .
O.12 J .HDX RRC a I.Ci.SIE(IKP) S]EB * SIE(]KW.LE.DIFFCO(1KL*i) DCL a D!FFCO(1KL+2) DCB * D]FFCO(JKL*3) ]PK )KP a a + K2NC * NWPC = 1K .IKBR) DO 84100 l=2.DASS(UC) AUL .EO.K2 DO 60000 13a1.0.25*RRC CYRX = CYLO0.KSR IK IX a I + (I1)*K2NC + (Ki)*NWPC + K .DIFFCC(IKL) DCT .2.U(JMKS) vW = W(IKMS) AUC .1 (12)(KER1) 84100 W(JK) a X(lX) C 89999 CONTINUE IF(ICONV.5(TS(IK)+TS(IPK)) . IBR(K2) + I ' (1i)K2NC IK = i + (K1)NWPC IKL a 1 + (11)K2NCL + (K1)NWPCL RC a FLOAT(I1)bDX .CALL TDM(A . 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 .NSEO) 33333 CONTINUE 15E0 a 0 DO 60000 K=Ki.1SP1 00 84100 K2.CF(]MKS) UC c U(IK) WC = W(!K) UL .5RRC*RDX CFC .D1.EO.S) SIECO .1CALI.NE.0P.X.I) GO TO 55000 DCR .1) GO TO 2040 ISEO = ISEO + 1 IF(ISEO.S]EO(]K) TSRA TSLA .SIE(IMKS) SIET .SIE(IPK) S*EL .NWPC CFR = CF(1PK) CFL .CF(UK) IF (CFC.K2NC ]K IK 7]KS IKMS = IK .5(TS(]K)+TS(lMKS)) VAR203G0 VAR20370 VAR20380 VAR20390 VAR20400 VAR20410 .DLeS(UL) ALC = DAES(WC) 1wS * DASS(WE) = SIE(1K) S!EC SIER .0/RC OR = CYL0. VAR20420.B.
(r'.OSrlth7A OV7 t!VA t. 10 + ( 0S669 01 00 (V03VIJfl S0..C*~ t )4* .± 4:7) 21J3 1u L7S1L'q'l OS 60ZMVA (Slx () I s I)SO I ((diE)l+ m1)l)'c = vzs I VIL .NOD (L03e."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. C0 r~t (r)LC 00~ttIdVA 060k ~wVA 0OLt dVA 00009 01 13t5.(N t0 *090~bVA 0O0I bVA 0O'0t MVA OSOI aVA 0OP0tZSVA (Q QtlY.~t J '~1f ((t0AX'1ZJANO ) C dt0* ) * N ) S 4~ I .L. N *:dM o zi Z)q4s)4 coo 0.x)j! 2(AI1)3rS oo0t 9 06CI ~1VA' 09ZI dVA 0LI.+c a 21J10 to.At~VA OSZI~7A clt'L fa ZAVtAZ'4 0=0i9 00 0=O9 00 I . 06C Zd7IA ZltrEzf 00 OOOL CO 0SCtLb7A NOI0t1D3JCZ NE d33MS CNZ 0L~tZZ7A LCECE OZ :tZ7A 1 05 3 3lNTI.1! Al z(IJ)L3 (NOfl'Ol V(ISJX017) Ot)t0dVA 0001ZlA N7 dlxiol ((L1)x~JNG4(IV~~sNQ)ts0 .Q ~ 000 fl co(3 S6 . (nN*31.I vo 0009 0:9>4 I )c t a. t :) a .OlI)03(r v+n . t7) .
5(DNFF2(3.(12) + K .B2.CO(IK)0.0 DI(J) D(J) + DIFFTDCT*RDZ GO 70 70000 65000 01( ) = SIE(]K) L2(j) a 0.11.EO. .OlFFTDCTRDZD IF(CFB EO.LE..TTSPLL .NE.CF(IKt S) WS W(IKMS) WC a W(IK) VAR21570 VAR21580 VAR21590 VAR21600 VAR21610 AWB a DLeS(we) AWC DOASS(WC) TSTA a 0.21730 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 .5(rS(lK)+TS(lKMS)) C GAT DIFFT DIFFB C C * RDTSIE(]K) .C2.RDZ C2(J) = 0.0 E2(J) .K2 Ix = KS.i) GO TO 70000 IF(CFT.NU) GAMT * RPRAN .5(DNFFZ(I.5 0.K)*DNFF2( .E0.ND.DIFFCO(IKL) = D]IFFCO(IKL+1) = DIFFCO(]KL+2) .NWPC C CFT = CF(IKP) CFB .DIFFEDCRD2D RDT * (DIFFTDCT+D)FFS*DC6)*A02D .(WC*AwCWS*iWB).X GtA.(A2.0 69950 IF(CFT.M * * IF(TS(]K).0 GO TO 70000 65500 C2(j) E82(J) .5.5(WCWC)... 1 * 0.EO.1.K2NC IKWS VAR21460 VAR21470 VAR21480 VA021490 VAR21500 VAR215 O VAR21520 VAR2rS30 VLR21540 VAR21550 VAR21560 = ]K .12 DO 71000 KKI.1) GO TO 65500 D1(J) * 01(J) .IF C DCR OCT DCL DC8 C ( CFC.0 70000 CONTINUE C CALL TDM.DIFFCO(IKL+3) 1PK a 1K + K2NC IKP a !K + NWPC IMKS a 1K .A2(J)SIE(3KMS) A2(J) • 0.IKBR) C DO 71000 ]=m1.I) )XMX GAMTTSTA VAR21620 VAR21630 VAR21640 VAR21650 VAR216GO0 VAR21670 VAR21680 VAR21690 VAR2I700 VAR217 tO VAR21720 VA.I DI(J) a TGA.0 C2(J) = 0.0.5"(ws*Aw )'RD2 .RDZ .5(TS(]K)+TS(3KP)) TSeL .k)+DNFFZ(1.K+1))X.C2(J)SIE(IKP) C2(J) = 0.0.X.01.NNOPT.1) GO TO e9950 Dl(J) = 01() .0.DIFFTDCTRDZD C2(J) a 0.K.1) GO TO 65000 .
12 DO 33335 K=KI.K2 IK v 1 + (1I1)K2NC + (K1)*NWPC STOR(I.IK r 1 + (II)*K2NC + (K1).09 KzK1.K2 1K 1 + (]i)'K2NC + (KI)*NWPC 33338 SIE(K) = 0.K) = SIE(IK) 333335 SIE(IK) .S)=W(KK+) TO(KMS)=TO(KK+1) VAR22470 VAR22480 VAR22490 VLR22500 VAR22510 VAR 22520 VAR22530 VAR225ZO VAR22550 .E0.1) GO TO 33333 GO TO 33339 IF(ISEQ.11.NWPC DO 2.NE. 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. TRANSFERS VELOCITIES TO STORAGE ARRAY( 2040 KI1 K2xKBP2 LWPC=I .NwPC 71000 SIE(]K) = X(IX) C IF(KIRK.O) 00 33335 I11i.K2 LWPC=LWPC+NWPC AT TIMEN ) . 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.5'( SIE(IK) + STOR(I.SIEO(IK) KIRK a i GO TO 33334 33337 CONTINUE DO 33338 1=1112 DO 33338 K=KI.K) ) KIRK a 0 33329 CONTINUE C C NOTE.
7. results are in Fig. 5. 5. The files. 'inputn dat'.1 and Fig.APPENDIX D TYPICAL INPUT FILES The file. The output The files.3(B). are for comparing Huh's And De Vahl Davis and Mallinson's correction formulas in the problem geometry. 4. The file. is the input to test the stability of the ADI scheme for momentum equation. 5. are for calculating the numerical solution by donor cell scheme in the problem geometry.8. 5. . 'inputd HUH' and 'inputd DM'.6 and Fig.2. 'input mom'. Fig. 5. 'inputd 3' and 'inputd 7'. The output results are in Fig. Fig. is for calculating the natural convection flow field in a Cartesian closed compartment.
0 2 l.0 1.0 0.9 0.4 0.0 1.01 0.0 0.0 1.0 0.0 0.0 10.0 0.0 1.0 4 0.0 0.0 2.4 5 1 1.0 0 0.0 1.0 0.0 0.0 1.0 0.0 0 0 1.0 0.120 1.t20 1.7 0.0 1.0 10.0 0.2 10.rTh FILE: INPUT MOM VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 4 4 O STABILITY TEST FOR MOMENTUM 2.0 0.0 0.0 0.0 1.0 0.0 60.0 2 5 10.0 .0 1.013 0.0 0.0 0.0 0.0 0.01 0 1.09375 0.0 0.0 1.045 0.5 6.0 1.0 0.0 0.44 0.0 0.0 0.0 4.0 0.0 60.0 0.4 0.0 0 2 1 3 O0 2 2 O O ADI 10.0 0.0 10.0 0.0 1.0 60.0 0.0 I 3 I 16G 1.5 0.75 10.4 1.
045 0.5 0.0 10.0 1.20 1.0 0.20 2 1.751.0 NATURAL CONVECTION IN A CARTESIAN CLOSED COMPARTMENT 1.20 II I 1.0 1.0 1.2 1.5 1.0 .0 0.0 1.0 1.01 1. 20 1.0 0.0 1.0 60.0 0 1.7 1.4 0 0 0 0 O 0 0 6 2 0.0 0.0 68.0 1.0 0.0 0.0 1.20 0.0 32.0 677.0 0.44 0.0 2.01 0.90 0.0 0.0 1.0 0.0 0 0 2 1.20 030 1.20 1.FILE: INPUTN DAT A VM/SP CONVERSATIONAL MONITOR SYSTEM 'PAGE 001 5 10 0 1 0.0 I 43 1.21795 1.+20 60.5 5 1.0 6 6 50.0 0.67 1 1I 2.0 4.0 0.0 8 2 I 3 1 3 0 3 0 1 1.
000024 0.0 1.CASE 3 2.0 0.0 1.0 60.063 ( ).0 1.02 1.063 0.FILE: INPUTD 3 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 10 10 0 0.0 0.911 3 4 1.70711 0.0 7.01 0.001 0 1.0 0.0 1.0 2 1.05.0 2 1t 1 113 5.0 0.026 5 1.457 1.0 0.000 9 1.000 10 1.44 4.0 0.0 0 0 NUMERICAL blFFUSION .0 1.0 0.0 5.O.204 0.000 0 0 0 O .0 2 4 2 4 1.002 0.75 0.014 ).009 G 1.47.0 12 11 2 121 t 5.0 .0 0.0 7.0 1.204 0.0 0.0 1.0 0.0 0.0 0.0 0.009 4 1.751 2 1.0 0.07107 7.5 1.0 0.974 5 1.0 0.3 1 I 1.0 0.940 1.0 0.07107 0.948 1.0 1.II I 11 2 2 1.07107 1.0 0.0 9.002 0.0 0.5 1.000024 60.045 1.0 1.0 0.0 0.0 9 10.0 0.0 1.0 10. 020 2.0 0.07107 12 12 Ill 2 5.0 7.0 1.0 1.0 0.0 0.0 0.01 0 0.000 0.OGI 2.07107 0.0 5.999 1.0 1.0 0.003 7 1.207 0.0 9.249 3 1.0 0.5 0.0 9.70711 1.497 0.9 0.0 GO.0 0.0 1 .0 0.0 1.O 2 1.0 0.000 10 1.0 1.0 8 10.0 0 2 5 3 0 7 6 1.991 6 1.511 1 1.0 0.014 0.0 5.09375 0.0 0.0 9.013 2.0 1.0 4 I 36 1.497 2.0 9.457 1.0 1.000 1.000 0 0 O O I.+20 I 0.0 0.07107 0.409 1.0 0.997 7 1.0 1.061 1 0.0 0.0 7.0 9.207 0.000 0.0 7.07107 111 3 1 1 2 5.0 1.0 0.
003 17051 1.0 0.5 1.0 0.0 5.0 4 2 4 1.0 1.001 0.0 0.0 0.07107 0.0 1.0 1.0 0.00.0 1.0 0.0 0.0 0.0 0.0 10.014 0.0 1 9.633 1.0 1.013 4 1.633 1.75 2.+20 2.0 0.17051 0.0 .0 0.0 1.0 1.0 3.0 10.0 60.014 0.0 1.071107 0.009 1.000 0 O 0 0 0 7 1 2 7 2 5.907 3 1.0 1.0 0.0 0.0 0.0 0.44 4.0 0.0 0.0 0.0 0.0 7.003 0.47.01 0 0.001 0.0.07107 711 8 8 2 5.0 0.333327 1.0 NUMER ICAL DIFFUSION .CASE 7 0.000 6 1.990 5 1.09375 .143 0.01 0.333327 60.0 7.3 1 1 1.000 0 0 0 1.01 2.000 6 1.0 0.0 7.FILE: INPUTO 7 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 6 0.0 0.07 07 0.002 5 1.457 2.4 1.07107 7.9 0.911 9.0 1.0 0.0 0.0 2 9.4 8 2 1 3 0 0.07107 I 113 7 2 1.0 1.0 0.0 0.374 1.0 1.013 2.0 5.167 1.143 0.0 0.0 0 0 2 2 1 0.0 1.0 7.626 1.0 0.0 I 1.0 2 2 0.0 1.0 4 1.0 0.5 1.0 60.457 2.045 1.0 0.0 0.0 2 0.0 1.0 7.07107 1 1 713 2 5.00.0 9.0 2 7 8 5.0 4 I 36 1.0 3 0.420 1.0 1.0 1.0 5.
0 0 *I O'l 0' *1II1 01l 0*1 0 1 0'L 01 O'1 0.0 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 ~W31.1 1' 0*0 0 *1 D99C00 0'1 0IO' 0*0 0*0 0*0 0'0 0 *I0*0 9c 1 t.0 09G"0 9Cs£0 661''0 6sv1'.*I riI 00 0*0 0*0 0i'00 0 10*0 6 v 0*0 0'0 top' . 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 .0 GLO10 ZU10 SL9*0 109.0 CGO. 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. 1'0 0.1bl b 690*6 £00*0 LC9 0 90c.z 10'o 0 1 0'l GL'O 1O1 '0 I 0*0 01l 0'l L910* 199*0 099.09 W '09 0.SAS 3~~~)Vd kOIINOW 0 0'0 0*0 0.0 0.. 0 LG0 .NOIS(fJJIG IV:)IUIW(IN too 39vd IVNOIIVSU3ANIX) dS/W~A v VIIi III)II OIMINI O~N :311A II . *0 PI 9.u *0 Ul' I 6P I 0011* I 0.0 9?LL '0 G'.09 0.L Z01 'L L'09C91C P09c91''? 1'Q9C9p 0'19C91' 1 iLO'O 1'LO*0 GL00O LLO'0 £00.0 01l 0'00' C0'I C.
20 911 1.0 0.0 0.0 2 25 25 5.20 913 1.20 I13 1.0 0 2 1.20 2 1.0 5.0 5.20 10 1.0 0.FILE: INPUTO IIU4U VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 2 0.20 1011 1.0 26 26 5.20 5.0 I 1 5.0 .0 8.20 9 1 1.6603 0.6603 0.0 2 25 5.20 1 1.20 2 1.6603 0.
0 1.0 0.0 1.120 2.016 0.0 1.0 5.246 0.013 2.96G 0.767 0.700 .0 1.0 0.0 1.0 1.0 0.0 5 1.0 0 2 1 3 0 7 6 I 0.0 1.636 0.044 9.197 0.0 1.0 0.0 2 2 2 0.5 1.460 4 1.0 1.003 0..0 1.003660 60.0 0.0 0.463604 2.0 0.0 0.0 60.0 1.030 0.156 0.302 0.0 0. 4G1IG04 7.4 4 1 36 1.951 0.003 0.0 0.0 I 1.4 0.0 0.0 0.0 0.5 1.0 0.30i9 1 0.029 0.011" 0.0 1.0 1.051 0.932 1.0 0.0 0.000 0.0 0.002 0.463G04 2.01 0 0.002 2.0 0.093 0.0 1.0 0.0 1.403604 2.420 0..0 1.02.0.0 0.045 1.0 0.02 2.4GI6 0.0 0.034 0.0 1.174 1.01 0.0 1.463704 9.705 0.0 1.0 5.0 0.099 0.072 ' II 12 13 14 15 16 17 10 19 20 21 22 23 24 0 0 0 I '.OGI 0.0 0.567 9.0 1.003 0.0 0.360 0. 1.0 0.0 1.0 0.0 60.026 0.0 0.20 1.009 0.496 0.GOG 2 1.526 3 4 5 G 2.33G 6 1.02.365 9.0 1.0 0.920 0.223 3 1.09375 0.5 1.0 0.0036G0 0.0 0.0 1.021 0.75 1.2406 1.0 1.133 10.0 1.0267 0.0 1.949 0.002 0.44 4.0411 0.005 0.144 1.903 0.0 0 0 NUMEtICAL IrFFUSION .076 0.522 0..0 1.rILE: INPUIO OM A VM/SP CONIVEIlSATIONAL MONI(fro SYSTEM PAGE 001 24 f( 0 I 0.566 0.0 1.242 7 1.030 0.0 3 10 1.0 0.0 2 4 2 4 1.260 0.3 1.463604 2.007 0.115 0.9 0.OM 0.969 0.
20 I 2 1.0 I*  .0 5.0 5.0 26 5.SILr.0 I 5.0 0.0 0.0 2 5. 6603 0.0 .0 0.20 113 1.: INP. & .0 2 5.6603 0.20 5.20 9 1 1.20 911 I.20 913 1.0 0 29 25 I 25 2 1.20 1. GG03 0.0 26 2 1.10 OM A VM/SP CONVERSATIONAL MONITOnr SYSTEM PAGE 002 2 0.20 1011 1.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.