This action might not be possible to undo. Are you sure you want to continue?
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS by
Kang Yul Huh and Michael W. Golay Energy Laboratory Report No. MITEL83011 August 1983
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS
by
Kang Yul and
Huh
Michael W. Golav
Energy Laboratory and Department of Nuclear Engineering Massachusetts Institute of Technology Cambridge, MA 02139
MITEL83011
August
1983
Sponsored by: Boston Edison Co. Duke Power Co. Northeast Utilities Service Corp. Public Service Electric and Gas Co. of New Jersey
2
ABSTRACT
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS by Kang Y. Huh and Michael W. Golay
A computer code is developed to predict the behavior of the hydrogen gas in the containment after a lossofThe conservation equations for the four coolant accident. components, i.e., air, hydrogen, steam and water, are set up and solved numerically by decoupling the continuity and momentum equations from the energy, mass diffusion and turThe homogeneous mixture form is used bulence equations. for the momentum and energy equations and the steam and liquid droplets are assumed to be in the saturation state. There are two diffusion processes, molecular and turMolebulent, which should be modelled in different ways. cular diffusion is modelled by Wilke's formula for the multicomponent gas diffusion, where the diffusion constants are dependent on the relative concentrations. Turbulent diffusion is basically modelled by the ke model with modifications for low Reynolds number flow effects. Numerical diffusion is eliminated by a corrective scheme which is based on accurate prediction of crossflow diffusion. The corrective scheme in a fully explicit treatment is both conservative and stable, therefore can be used in long transient calculations. The corrective scheme allows relatively large mesh sizes without introducing the false diffusion and the time step size of the same order of magnitude as the Courant limit may be used.
A
3
ACKNOWLEDGEMENTS
This research was conducted under the sponsorship of Boston Edison Co., Duke Power Co., Northeast Utilities
Service Corp. and Public Service Electric and Gas Co. of New Jersey. support. The authors gratefully acknowledge this thank all those who assisted them
The authors
during this research especially Dr. Vincent P. Manno, Mr. B.K. Riggs, Ms. Rachel Morton and Prof. Andrei Schor. work was expertly typed by Mr. Ted Brand and Ms. Hakala. Eva The
2.1.1.1 4.3 Low Reynolds Number Flow CHAPTER 4: NUMERICAL DIFFUSION Truncation Error Diffusion 4.1 Continuity and Mass Diffusion Equations 2.4.2 Physical Constraint 4.2.2.2 Truncation Error Diffusion 4.2 Energy Equation 2.3.3 Basic Assumptions and Limitations 2.2 Turbuletit Diffusion 3.3 Truncation Error Diffusion in a Two Dimensional Problem 4.1 Assumptions 2. Implicit and 4.3.1 Diffusion Convection 4.1 Derivation of Turbulence Equations 3.5 SMAC Scheme and Compressibility CHAPTER 3: DIFFUSION MODELLING 3.4 ADI Schemes Conservation 4.1 Molecular Diffusion 3.1.2 Governing Equations 2.1 Skew Differencing by Huh 4.4.3 Discussion of Skew Differencing and Corrective Schemes Raithby's Scheme 4.4.1 4.2.3.1 Truncation Error Diffusion in a One Dimensional Problem Another Approach for a One Dimensional 4.4.3.4 Solution Scheme 2.2 Comparison of Explicit.2 Crossflow Diffusion 4.1 Mesh Point Implementation Mesh Interface Implementation 4.4.4TABLE 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 2.2 ke Model 3.3.1 Problem Statement 2.4 4.3.2 .2.2.3.3 Tensor Viscosity Method Corrective Scheme by Huh 4.2 Limitations 2.3.2 4.4.
5
Page 4.4.3 Accuracy 4.4.4 Stability 4.4.4.1 Explicit Scheme 4.4.4.1.1 One Dimensional Central Differencing of Convection Term 4.4.4.1.2 'One Dimensional Donor Cell Differencing of Convection Term 4.4.4.1.3 Two and Three Dimensional Donor Cell Differencing of Convection Term 4.4.4.1.4 Stability Condition in Terms of Cell Reynolds Number 4.4.4.2 Implicit Scheme 4.4.4.3 ADI Scheme CHAPTER 5: RESULTS . . . . . . . . . . . . . . . . . 91 93 94 94 98 101 106 113 117 119 119 119 123 123 125 126 126 132 132 138 158 164 164 164 . . . . 168 168 168 .. . . . 170 176 181 184 204
5.1 Natural Convection Results 5.1.1 Updating of the Reference State 5.1.2 Energy Convection 5.1.3 Heat Transfer Modelling 5.1.4 Turbulence Modelling 5.2 Numerical Diffusion 5.2.1 Truncation Error and Crossflow Diffusion 5.2.2 Validation of Huh's Correction Formula 5.2.3 Skew Differencing Scheme 5.2.4 Corrective Scheme 5.3 ADI Scheme CHAPTER 6: 6.1 6.2 CONCLUSION . . . . . . . . . . . . . . .
Physical Models Numerical Schemes RECOMMENDATIONS FOR FUTURE WORK .
CHAPTER 7: 7.1 7.2
Physical Models Numerical Schemes . . . . . . . . . . . . . . . .
REFERENCES APPENDIX A: APPENDIX B: APPENDIX C: APPENDIX D:
ANALYTICAL SOLUTION FOR THE PROBLEM IN FIGURE 5.3(B) . ............ ADI SCHEME.FOR MOMENTUM EQUATION . . . . . . . . . . . . . . . COMPUTER PROGRAMS TYPICAL INPUT FILES ...........
6
LIST OF FIGURES Number 4.1 4.2 Illustration of the crossflow diffusion in a single mesh with pure convection ... ... Illustration of the crossflow diffusion in a row of meshes aligned in the xdirection with pure convection . ........ Geometry for explanation of the crossflow diffusion in a two dimensional . . . Cartesian coordinate ...... . Page 63
65
4.3
...
.
.
66
4.4
Geometry for explanation of the crossflow diffusion in a rotated two dimensional Cartesian coordinate . . . . . . . ....... Comparison of the prediction formulas for crossflow diffusion constant by De Vahl Davis and Mallinson and Huh . . . . . . . . .
.
.
68
4.5
. . 71
4.6
Geometry for explanation of the crossflow diffusion in a three dimensional Cartesian .. coordinate . ......... ...... A segment of the calculation domain showing grid lines, control volume and notations in Raithby's scheme . ........ Three possible flow configurations for mesh interface implementation of the corrective scheme in a two dimensional ....... .. ..... . . ... problem ...
.
72
4.7
. 80
4.8
.
85 92
4.9 4.10
Profile assumptions of 0 in space and time for various differencing schemes . ....... Three distinct shapes of the parabola, f(t), according to the coefficient of second order term ................. Stability condition in the plane (C xd x ) for an explicit scheme with central differencing of convection term in a one .......... dimensional problem . ... Stability condition in the plane (C ,d) for an explicit sc'heme with donor cell differencing of convection term in a one dimensional problem ..............
. . 97
4.11
.
. 99
4.12
.102
7
4.13
The region in the complex plane where (al+a2 ) should exist for the stability of an explicit scheme with donor cell differencing of convection term in a general two dimensional problem .............. The region in the complex plane where the amplification factor, r, should exist for the stability of an explicit scheme with donor cell differencing of convection . term in a general two dimensional probl Stability condition in the planes (C ,d ) and (C ,d ) for an explicit scheme wfthx donor elY differencing of convection term in a general two dimensional problem .
.104
4.14
. . . 105
4.15
.
. .107
4.16
Stability condition in the plane (C ,d ) etc. for an explicit scheme with doiorx cell differencing of convection term in a general three dimensional problem . ..... . .108 Stability condition in the plane (Rx,C ) and (Ry,Cy ) for an explicit scheme with donor cell differencing of convection term in a general two dimensional problem . .. .111
4.17
4.18
Stability condition in the plane (Rx Cx) for an explicit scheme with donor cell differencing of convection term in a two dimensional problem with f=v/ u/x.... .. 112
4.19
Stability condition in the plane (Cx , d x ) for an implicit scheme with donor cell differencing of convection term in a one dimensional problem .............. Stability condition in the plane (R xCx) for an implicit scheme with donor cell differencing of convection term in a one ........... dimensional problem . ... Natural convection flow field due to heat transfer between the obstacle and air in the containment at t=10.8 sec . .
.115
4.20
.116
5.1
.
. .
. 120
8
5.2
Natural convection flow field due to heat transfer between the obstacle and air in the containment at t=62.5 sec . ..... Problem geometry for the two cases, (A) parallel flow and (B) cross flow, for evaluation of numerical diffusion . . . Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 10x10 meshes along the line CD in Fig. 5.3(A) . Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CD in Fig. 5.3(A) . Comparison of the analytic solution, analytic solution with increased diffusion constant and numerical solution with donor cell differencing of convection term in 10xlO meshes along . the line CA in Fig. 5.3(B) ........ Comparison of the analytic solution, analytic solution with increased diffusion constant and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CA in Fig. 5.3(B) . . . . . . . . .... Problem geometry with arbitrary angles of e and 81 to compare De Vahl Davis and Mallinson's and Huh's formulas for prediction of crossflow diffusion constant . . . . . . . . . . . . . . . . . Comparison of the numerical solution with donor cell differencing of convection term and analytic solution with increased diffusion constant, (D+Dcf), by Huh's formula along the line BD'in Fig. 5.8 ... . . . . . . . . . . . .
121
5.3
127
5.4
128
5.5
.129
5.6
. ..
130
5.7
. 131
5.8
. . . 133
5.9
134
..convection problem with 6 = 450 Comparison of the true solution and solutions by donor cell scheme. .15 . . problem with 6 = 63.14 140 5. .3(B) . 136 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.3(B) . numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig. . . .. .13 139 5..3(B) .. . . CA in Fig.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 . . .17 Comparison of the analytic solution. 5.16 143 5. . . . 137 5.. . . 142 5. 144 . .. . . by De Vahl Davis and Mallinson's formula along the line BD in Fig.430 . 135 5. . . True solution and donor cell solutions with and without diffusion correction for a pure convection . skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with 6 = 63. (D+DDM). . Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 6x6 meshes along the line .. . 5. Donor cell solutions with various diffusion corrections for a pure convection problem with 8 = 450 . . 5.10 Comparison of the numerical solution with donor cell differencing of convection term and analytic solution with increased diffusion constant.12 . . . . 5.9 5. .43* .8 ..
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.20 .18 Comparison of the analytic solution. .. 152 5.10 5. . .. 145 5. 146 5. .. .. . .. . .. ... . .. . .19 . .. . . . .. Comparison of the true solution.. 5.. .. . 5..25 Diffusion constant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment . . . . .. .. 151 5.. .. .3(B) . .23 . Simple recirculating flow field for test of the implementation strategies of the corrective scheme .19 . ..19 .22 150 5. .. . ... . numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig.24 . . .. . 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.21 S. 5. . 148 5.147 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 ... ... . .. 153 .
. . . Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig.27 . 0...0 sec . . Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. with mesh point implementation of the corrective scheme . .0 sec along the line AB in Fig. . .3 sec. .1I 5. with donor cell differencing of convection term without any diffusion correction . . .30 with .. .. .. 0. . . . 0. .0 sec . .3 sec..30 5. . with mesh interface implementation of the . .. . 154 5. . . . Steam concentration distribution after one time step. .. . .3 sec...0 sec and the Courant limit of 1. . 155 5.31 1i60 5. 157 159 5. .. . corrective scheme . 156 5.29 Steam concentration distribution after one time step. .. . . . . . 5.cx T where 4 is given by. 163 A.30 for At = 0. 5. .5 sec and the Courant limit of 1.30 for At = 1.26 Diffusion constant corrections of the mesh interface implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment . . . .. . .. Problem geometry for comparison of the explicit and ADI schemes . Steam concentration distribution after one time step.28 .33 Comparison of the ADI solutions for the profile of 4 at t = 10.. . . . 5. .3(B) = e.32 161 5. . different time step sizes Boundary conditions in terms of 4 for the problem in Fig. 5. .. 1 178 . .
donor cell differencing of the convection term .2 . .114 . .12 LIST OF TABLES Number 4. . . . . . . 2D. . . . . . . . . . Comparison of the Von Neumann analysis and numerical experiments for the stability condition in terms of the cell Reynolds number in an explicit. . Page .1 List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research. . . . 1981 . . . . 78 4. . . .
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.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 . cp/c Length Mass Molecular weight Nusselt number Peclet number Effective Peclet number (1) Pressure u u (2) L* ax or Ax Ay Ax + p v C .C x y D D ah D as Dhs D .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.D y.
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 .1 v u Mesh configuration.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. 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 .
Some simplifying assumptions are made concerning the thermodynamic state in the slow mixing stage.I~~U1i* 11*11044~1 1~. hydrogen. the fluid dynamic phenomena in the containment in order to justify those remedies. it is necessary to understand such problems. 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. air. the fast blowdown stage and the slow mixing stage. The distinct feature of the second stage is a much longer time scale in comparison with the first blowdown stage. although it may be less than 100% when there is no liquid component present. The four components. Some remedies have been proposed to deal with However.__ 16 CHAPTER 1 INTRODUCTION One of the major concerns in the Three Mile Island (TMI) accident was hydrogen gas accumulation in the containment. The error due to decoupling is negligible in a slow transient where the . The governing conservation equations are decoupled in order to simplify the solution procedure.( 1~111111_~. steam and liquid droplets are assumed to be in thermodynamic equilibrium and the relative humidity is assumed to be 100%. the response of the containment during an accident can be divided into two stages. From the hydrogen transport point of view.
17 state change over one time step is small. molecular and turbulent. are determined by their own transport equations. resulting in the same convection velocity for the four components. Convection and diffusion are the central issues in physical modelling efforts of hydrogen transport. and the total diffusion constant is the sum of the diffusion constants of those two mechanisms. In the first step the continuity and momentum equations are decoupled from the energy and other scalar transport equations and solved by the Simplified Marker and Cell (SMAC) method in order to obtain the flow field. The molecular diffusion constant is predicted by Wilke's formula [73) for multicomponent diffusion and ChapmanEnskog formula [5) for binary diffusion. c. the phase change is taken into consideration to maintain 100% relative humidity. Convection is assumed to occur as a homogeneous mixture. and turbulent dissipation rate. . The turbulent kinetic energy. Diffusion occurs by two independent mechanisms. k. Finally. The turbulent diffusion constant is predicted by the kc model [45). 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. The diffusion constant. of each component in multicomponent gas depends on the mole fraction of that component. The turbulent kinematic viscosity.
Truncation error diffusion is a one dimensional profile error and crossflow diffusion is a multidimensional operator error [68) of the finite difference equation. the major obstacle in accurate numerical modelling of convection and diffusion has been the crossflow diffusion error which arises in donor cell treatment of the convection term in multidimensional problems. The leading issue in numerical modelling of convection and diffusion is to minimize the error that occurs in the numerical solution procedures. Since the error usually appears as an additional diffusion. The numerical diffusion has two different sources. This is because the gradient of the scalar quantity under consideration is small in the flow direction in comparison with that in the direction normal to the flow. numerical diffusion. Therefore. There has been much debate on the numerical diffusion and many schemes have been suggested for the past two .18 which is the diffusion constant for momentum transport can be calculated directly from k and E. The turbulent Prandtl and Schmidt numbers are assumed to be equal to one. Truncation error diffusion occurs in the flow direction while crossflow diffusion occurs primarily in the direction normal to the flow. however the latter turns out to be the dominant error source in most convection dominant problems. has been used to describe the numerical error in general. truncation error diffusion and crossflow diffusion. the term. The effective diffusion constants of the two errors are of the same order of magnitude.
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. although with different stability conditions. Raithby's [53) and S. The corrective scheme is inherently better than the skew differencing scheme in that it is conservative and does not affect the simple solution procedure. a recent review paper on this topic [65) shows that there is no scheme universally acceptable consistent with using reasonable mesh spacings and computation times. The corrective scheme can be implemented with any of the explicit. have been tested for recirculating flow problems. mesh point and mesh interface implementations. skew differencing and corrective 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. 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. However. The schemes presently being used can be divided into two categories.19 decades. Two implementation strategies for the corrective scheme. ADI and implicit schemes. The conservative property is essential in a long transient problem like the hydrogen transport in the containment.
simulation results compared with experimental data. The suggested physical models and numerical schemes have been used to simulate the LOCA experiments performed in BattelleFrankfurt [44) and HEDL (Hanford Engineering Ref. . [49) includes all the Development Laboratory) [6].20 implementation has always given physically reasonable solutions and may be used extensively for all diffusionconvection problems.
a large amount of steam and water will come into the containment in the fast blowdown stage increasing the containment pressure. The major safety concern is that of keeping the containment pressure below a certain level to prevent a large scale leakage of radioactive materials..g. installation of ignition devices. Thereafter. . it is essential to understand the fluid dynamic phenomena in the containment.21 CHAPTER 2 PROBLEM STATEMENT AND SOLUTION SCHEME 2. containment inerting.. for dealing with this problem. There have been some mitigation procedures suggested. In addition to the pressure increase due to the primary coolant. 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.1 Problem Statement Once a lossofcoolant accident (LOCA) occurs in a Light Water Reactor (LWR). After a while the slow mixing stage follows the initial blowdown stage and continues for an extended period of time. e. 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. use of flame suppressants and enhanced venting capability. the hydrogen has received much attention in the safety analysis of nuclear power plants.
thermal nonequilibrium.g. some simplifying assumptions should be made to use the numerical procedure without impairing the acceptable solution accuracy. laminar and turbulent flows. phase change and heat transfer between the gas components and wall. . Therefore._ __ _ __ _ I_ ~ 22 Since a numerical method is suitable for this purpose. e.. The most difficult aspects of this analysis are the complicated geometry and chaotic postLOCA conditions. a good computational tool has been required to predict the hydrogen concentration distribution. etc.
S + + t Bui P au k aui + ax~ ) x k X axk gk Pr t t 'T xk (2.5) Turbulence dissipation rate: D 1 _ 1t StE xk Dt P Txk 0E ] + C t*( 1 p k ui xk ouk au.Vp i + 4.6) Equation of state: p = f p (ii .4) Turbulence kinetic energy: Dk 1 t 3k 3 P 1 k ok x .T) (2.2 Governing Equations The governing conservation equations are set up to describe the postLOCA fluid dynamics in the containment. Continuity: V'v= 0 pl + Vvv ] = Vp + Vv = 0 (2. The physical implications of the conservation equations will be given with their basic assumptions and limitations.7) .xk k 2 v t BT C2 + kg k Pr 3x 2k k t (2.2) Energy: p[e e + Vvh] + = VkVT + e fg (2.t = VD.1) Momentum: pgI + d + V0 (2. (2. I+ 3x 1 .23 2.3) Mas s diffusion: S+ V*vp i .
T) (2.12) (2.16) kt = . Im + Dit it (2.18) p = Epi (2.24 Constitutive equations: 'N = f (p .T) (2.14) m = fk (PiT) (2. .19) Note that 1. = D. = f (Pi.13) D.11) k = .t (2.10) v = vm + Vt (2.17) t = C k 2/C E (2. The subscript i denotes the four components as .9) D.k + k t m Dim = fD(Pi'T) (2.8) f (v) (2.15) Dt = v t (2.
v. three momentum. w. Therefore. i=14. The mass diffusion equations are summed up for the four components.25 follows: i=l i=2 i=3 i=4 2. H2' steam.21) D4 = 0 (2.k C for a three dimensional case.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. air. u. four mass diffusion and two turbulence equations with ten primary unknowns. T. 3. The turbulence equations are written in terms of the Cartesian components in tensor notation.i (4). there is one moreequation than is required. 2. liquid droplet. . (2. one continuity. The phase change occurs between steam and liquid droplets and the diffusion constant of liquid droplets is equal to zero.20) i = 2 = 0 43 + 04 = 0 (2.22) There are eleven conservation equations. one energy. p.
Therefore. 2.26 are satisfied. 2. 2. Vp = 0 (2.) i + E 1 (2. 2. 2. 2.Vp.26 . Eq.1 and Eq. 2. it is reasonable to assume that Eq.) + V(Ep Jt i = (V'D.25 and Eq.23) The sum of the phase change terms is equal to zero.27) Consequently Eq.27 remain valid throughout the transient.27. at p= 0 (2.26 can be reduced to the following Eq. Since the concern is only for the slow mixing stage after a LOCA.(Ep i ) a .26) It is a reasonable assumption in a turbulent flow regime that the four components have the same diffusion constant. ap at S+ Vp = E(VDiVp i ) 1 (2. . 2.24) Now it can be seen that the continuity equation is valid if and only if the following Eq.25 and Eq.25) Vp i = 0 (2.4 are consistent only if the temporal and spatial variations of the total density are very small.
(p. [c pihi .2 Energy Equation The energy conservation equation.3. (2.31 so that the internal energy and enthalpy can be written in terms of the temperature. 2.2.iei) + V'[Ep. is also of an approximate nature and the exact form is given in the following. Eq.29) The second term on the left hand side of Eq. p + £e. E it p + e.27 2.i = V'kVT (2. LeaPi LeQ. I at t + V [vp.30 is substituted in Eq.29 denotes the energy change due to the concentration change in a given control volume.31) The thermal equilibrium assumption is incorporated in Eq.2.ih i ] = V*kVT (2. ae.30) Eq. It is clear that the major contribution to this term will come from the latent heat of the phase change.i + V.ih. 2. 2. Be.29 to give the following. .28) The temporal term can be divided as follows. 2. = VkVT (2.
2.32 become identical by defining average internal energy and enthalpy as follows. P (2 33) (2.T h.3 Basic Assumptions and Limitations The governing equations are based on some simplifying assumptions about the physical phenomena in the containment after a LOCA. Pic vi T c p p.3 and Eq. = c p1 .33) h (2.1 = c Vl.28 (Picvi )~ +V"[vTZpic pi] + efg 9kVT = (2.32) where ei i = eg4g + e 4) Sefg g e. 2.T P P e Piei p ih p.c .T 1 Eq.34) 2. It is necessary to clarify these assumptions and their limitations for a safe use of the given physical models and solution scheme. .
2 Limitations 1. The relative humidity in the containment is 100%.29 2. The change of the diffusion constants. with no slip. Dt and k t over one time step should be small because the turbulence equations are decoupled from the velocity field calculations. 2. The four components are in a state of thermodynamic equilibrium. The turbulent Prandtl and Schmidt numbers are equal to one. vt. 3. The temporal and spatial variations of the total density should be negligibly small because of the incompressibility assumption. The four components are treated as being nearly incompressible and the phase change rate is moderately small.3. . 2. The four components all have the same convection The liquid droplets move with the gas components velocity. The relative humidity may be less than 100% when there is no liquid component left. 5.1 Assumptions 1. 3. 2. 4. The phase change rate should be moderately small because the solution scheme decouples the phase change term from the energy conservation equation.3.
Pi(4).7 are a coupled set of equations with primary unknowns. p. The equation of state.30 2. E. explained in detail in section 2.4 Solution Scheme Eq. . which are the four mass diffusion equations. energy equation and two turbulence equations. The SMAC scheme is In the second step. In the first step the SMAC scheme is used to get the convergent velocity field with zero divergence for every mesh. 2. the relative humidity may be less than 100%. is used to update the reference pressure and reference density which is important in calculating the buoyancy force. It solves the .7.5 SMAC Scheme and Compressibility The SMAC (Simplified Marker And Cell) scheme [21J has been used for the incompressible fluid flow with Boussinesq approximation for the buoyancy force. the obtained velocity field is substituted in the scalar transport equations. the conservation equations are decoupled into two sets.5. 2.i Since it is too difficult to solve the whole system of the equations and obtain a consistent solution for all the unknowns. . 2. T. The phase change is assumed to be zero in this step. In the third step phase change is taken into consideration to maintain 100% relative humidity. u. continuity/momentum equations set and other scalar transport equations set. Eq. k. v. When there is no liquid component left.1Eq. 2.
However.e. 2. i.37) Py The vorticity conservation equation is derived from Eq. ay (Eq. which can arise with net inflow or outflow boundary conditions.36) .36) ydirection momentum: av av ua av 1 + pg + a y x2 Nv 2 (2.. The continuity and momentum equations are given in two dimensional Cartesian coordinates in the following.37 as follows.35) xdirection momentum: au au + u+v at ax au ay 1 p ax + Pgx + a 2u 2 ax 2 2au + 2 2 (2. (Eq. Continuity: au ax av ay 0 (2.37)y (2.1_ _~ 31 continuity and momentum equations to get the velocity and pressure field in an iterative way. it is possible to modify the SMAC scheme to accommodate a slight compressibility effect.ax 2.36 and Eq.37): . the divergence of the velocity field vanishes everywhere. 2. the SMAC scheme can be used only for incompressible flow calculations. 2. Since the continuity equation used in the SMAC is an incompressible form.
38) =ay ax The Poisson equation for the pressure field is also derived from Eq. In the tilde phase the approximate velocity field at time step (n+l) is obtained from Eq.37): a2 = 2 ax 22 axay a 2 ay 2 (a2D + a2D aD +1 2 at Re ax 2 ay (2. 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. .+ y ax ay The basic idea of the SMAC scheme is to separate each calculational cycle into two parts.40) S (2. 2.41) where fn and gn are the explicit quantities at time step n. 2.x ay ax 2 ay2 where au av (2.36) + ~(Eq. the socalled tilde phase and pressure iteration.39) where av au D= .37 as follows. ax 2 2. 2. (Eq.36 and Eq. ay 2.32 a C 2 a2 + u+ V = v+ ) at .37 as follows.36 and Eq. 2.
2.45) .44 is given in the following.42) av . shows that the vorticity is independent of the pressure field.43) ax ay It can be seen from Eq. 2.41 have the right vorticity.40 and Eq. follows.41 are still unknown.39 may be rewritten as A finite difference form of Eq. 2.33 Since the pressure field in Eq. 2.v) D (2. the following relations hold after the tilde phase. 2. a guessed pressure field or the pressure field at time step n is used to start the iteration. 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. 2. _ Pi+j2P +P iPij+ pij f(uv)  Dn + l D n D D Ax 2y t (2.40 and Eq. The equation of vorticity transport. 2.38. 2.44) Eq. au ay au av ax  (2.O 0 (2. The impor tant point is that un + l and vn+l calculated from Eq. V2p = f(u. Therefore.
equal to zero.49) . sidering the cell (i. 2. (6u) L =  pAx (6u) R = t6p pAx (2. no more correction will be made.46) (2. Once the zero divergence velocity field is obtained. 2.37 so that it does not change the vorticity implemented in the tilde phase.j) 26p( 1 Abx 6p = h + _) Ay 1 At( A + 1 Ay 1 Dt D (2.47) ) The pressure correction is given in terms of the residual divergence for every mesh.. The velocity correction formula is derived from Eq.48) The SMAC scheme can be extended to a slightly compressible fluid flow.34 The pressure correction formula is obtained by conn+l only and putting D. The slight compressibility means that the NavierStokes equations may be solved safely with the incompressibility assumption. while the continuity equation has the following compressible form.36 and Eq. S+ at V"(p) = 0 (2.
50) The two terms on the right hand side of Eq.35 From Eq. : in n =Vp v  total net inflow of mass at the boundary during the time At. Now the problem is how to converge to a predetermined nonzero divergence field. D = V'v 1 p p at V p (2.50 may be given from the previous time information and boundary conditions. 1 .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.p_ p at 1 min/V At n p n n (2. 2.49 the nonzero divergence may be derived as follows. 2.52) (2. The associated physical assumption in Eq. 2.51) v p where m.uniform over the whole containment. 2. .51 and Eq. V: total volume of the containment. This can be done with the following modification of the pressure correction formula of the SMAC scheme.
(DD 1 1 + ) 2At( Ax Ay (2. The velocity correction formula remains the same as Eq. 2.53) where D O is given by Eq.48. 2.6p = .50. .
air and steam. air. Convection occurs as a homogeneous mixture resulting in the same convection velocity for the four components. 3. hydrogen. molecular and turbulent. Diffusion occurs by two different mechanisms. Although turbulent diffusion is greater than molecular diffusion by a few orders of magnitude. proper modelling of the latter is important because the molecular diffusion of hydrogen is significantly greater than those of other gases and also because there may be a laminar flow region in the containment.1. The thermal conductivity and viscosity are also diffusion constants in a broader sense for momentum and energy transport.1 Molecular Diffusion Molecular diffusion occurs by the collisions of gas molecules. which should be modelled independently. in the containment after a LOCA. the diffusion constant of each component is calculated separately by Wi~k's formula [73] in Eq. 3. It gives the diffusion constant of each component in terms of . As collisions occur more frequently. steam and liquid components are transported by convection and diffusion in the containment. components.37 CHAPTER 3 DIFFUSION MODELLING The hydrogen. the process Since there may be three gas of diffusion is also increased.
: yi: Diffusion constant of the ith component Mole fraction of the ith component D. 3.2.2) poAB D. The property of air will be replaced The resulting diffusion con with that of nitrogen gas. stants are given as follows.38 the binary mixture diffusion constants and mole fraction of that component. T3 D = 0. The binary mixture diffusion constants are calculated by the ChapmanEnskog formula.0018583 1 + M A MA MB (3..: Binary mixture diffusion constant between the ith and jth components. Eq.AB where DAB: AB Binary mixture diffusion constant in [cm /sec] .1) where D. 1Y A B DAB 1Y B yA DBA 1Y C YA CA C YC CB B YC DBC C D = C YB DCB (3.
P n ].UIIII ~ II~PI^ICIL1I n~s^  ^ 39 T: p: aAB: MAMB : Temperature in [OK] [atm) Total pressure in LennardJones parameter in [A] Molecular weight QD.3) CAB = AB For the pair of polar and nonpolar gases a correction factor is introduced to account for the polarity.4) np = np 2 where the correction factor 4 is given by. =n [ p n = [+ S n .2.AB: Dimensionless function of the temperature and intermolecular potential field for one molecule of A and one of B. ap = np £ (o n + o p) EE e (3.3. 3. (3.lll~*. In order to calculate the para meters aAB and 0DAB in Eq.5) 2 n . 3. For nonpolar gases OAB and ODAB are calcu lated by Eq. The error range of the ChapmanEnskog formula is reported to be about 610% [5).. OAB = (o A + oB (3. the following relations are required.
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. 5 = 4.65 A The polarizability is given for the hydrogen and nitrogen (air) components.5 are defined as follows. The containment pressure is assumed to be1 atm.2 E/K = 380 0 K a = 2. 3.= 1.9 x 10 25 25 25 [cm3 3 [cm ] : N2 : an = 17.6947 as .9035 = 4. H2 a n = 7.6 x 10 The fitting functions for the binary mixture diffusion constants are obtained by the leastsquarefit method. t* .9492 x 10 D ah D 1 T .40 The parameters in Eq.6 T1.6810 x 10 .
6) The viscosity and thermal conductivity can be considered as diffusion constants for momentum and energy transport and calculated similarly as follows [5).8916 x 10 where T: D: [OK1 [cm 2 /sec] 5 T 1.10) Viscosity in [poise) Temperature in [K]) LennardJones parameter in [A) Dimensionless nTumber . i (3.8144 (3.67 x 10 where : T: a: Svi : o2 2 (3.lllI_ 41 Dhs = 2.__Li~n_~PIIII ~.k.7) k. n mix = Yipi n i E j=l Yj ij (3.8) j=l where ij 8 Mj j ij M The viscosity of nonpolar gases may be obtained by the following [5).lii. = 2. = n i=l y.
[5]. 3. The . c (3. Wsteam = 0.2 Turbulent Diffusion The diffusion process is greatly enhanced by the turbulence of fluid flow.12) where c P is the specific heat per mole at constant pressure. and turbulent dissipation rate.42 The viscosity of steam is.385 nn y. There are several models suggested for Presently the best model calculating turbulence effects. E.009 cp at 1 atm. The ks model sets up the transport equations with some empirical constants for the turbulent kinetic energy. 200 C There is another formula suggested for the viscosity of multicomponent gas in the following [5].11) for the Prandtl number of Pr = c (3. 2 n Yi umix = C 2 i=l Yi + 1. seems to be the kc model originally developed by Launder and Spalding [45). k. The effective diffusion constant is therefore the sum of the molecular and turbulent diffusion constants.y RT k=l PMiDik k i There is a useful relationship nonpolar polyatomic gases.
13) (3. (3.' For example. u. 34. + u. These terms are treated by the eddy viscosity concept of Boussinesq and eddy diffusivity concept which are given in the following. 3. momentum and energy equations by decomposing the variables into mean and fluctuating parts and averaging the resulting equations. The effect of turbulence appears as the additional terms due to the product of fluctuations which do not necessarily cancel out.14) .' j T'= a t x.= v ( Dt + ui j u P= Dt ax x 2 k6 ij 3x_) 3 i (3. 29.u.43 turbulent viscosity is given directly in terms of k and C.u.' etc. u. . etc.15) . + Pi Pi = Pi + p.2. = u.1 Derivation of Turbulence Equations [15. 45] The transport equations for the turbulent kinetic energy and turbulent dissipation rate are derived from the continuity.
A two dimensional cylindrical coordinate form of the kE model is introduced in Eq.44 The transport equation for u.2 ke Model The transport equations for the turbulent kinetic energy and turbulent dissipation rate form the basis of the kE model. j=i.' u is transformed to the turbulent kinetic energy equation by the contraction. 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.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 ] .16.c +g t Pr v BT az Tt (rue) + + 1 r ar a(w)= z [) [ pr ar o ar + z a ( a az + C 1 aw [ 2 au 2 t + 2() 2 () p az ar p k 2 2 + 2 2 C + ( r + ) az 2 2 k + k Pr atT zt (3.2. 3.
C 0.45 The turbulent viscosity pt is given in terms of k and E in Eq. Since the kE model is applicable to fully . 3.0 o 1.3 The molecular effects can be included in Eq.44 C2 1.17) The suggested values of the constants are given in the following table. it ak t a +it m ok lit m a E C The turbulent Prandtl and Schmidt numbers are usually assumed to be equal to one.3 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. 3.17. = C pk 2/ (3. it is difficult to determine whether a region under consideration is in a turbulent or laminar flow.16 as follows.92 ok 1. 3.09 C1 1.metry.
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 2.5/(l+Rt/50)] C2 =2 where R 2 0 [1. Eq. 41). Rt.16. it is required to extend the model to laminar flow regime or to switch to laminar models according to some criterion. 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. 3. The following kc model is a modified form by Jones and Launder [40. CP = CP0 exp[2.3 exp(R2t ) (3.lit k + l ak aui i + t (ax k + xk auk axi axk e .46 turbulent flow.2vak 2() k 2 (3. while C1 and C2 are to vary with turbulence Reynolds number.0 .19) = pk2/VE . The model presently being used is just to use the sum of the molecular and turbulent contributions for every region in the containment.0 a ui ( 2 p ax ax 2 ) Dk Dt 1 p ak [(.0.
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)
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.3.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.a(?) ) 2(.6 from Eq. 1 < c < n 2 4 + i Ax 2 11 4 2 1 i+d < d < 1 The mean value theorem.y+b].i+c T i+d ax where the following relations have already been used. . we obtain the error at time step (n+l). n+g 4 i+ Ax n i Ax 1 xji+ n i+l .51 Subtracing Eq. 4. 4. 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.x+a) and [y.y) + (f) af (4. It can be extended to a two dimensional form as follows.7. n . is for a one dimensional case. af f(x+a. Eq.8) 3 @n + u()n + a( ax i+b ax 4 n+g _ n )n+g .y+b) = f(x.
52 Now Eq. 4.10) The derivatives with respect to time and space in Eq. 4.C ax 22 +At g t .F (4. C and E are truncation error diffusion terms while the terms B. D and F are time derivative terms which become negligible for slow transients.tax . . The terms A.10 are based on the assumption of perfect information at time step n.D +aAx(cd)Atax 2 +agt . 4.9 reduces Eq. 4. 2 n+l (4i ) EXACT  (4+) FD = i n+1 f )2 2fx 1f 2 ) (f 2 2 .xi+ ] and [tn tn+). 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 . E . they may represent the errors over many time steps in a stable scheme.8 to the following form. Although the error terms in Eq.A fAxAt axat 2 B utx(a+b) .
4. The analytical solution of Eq. cL (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 4.1.53 4. 4= Cecx + C2 where c = u/a .11) ax Consider a case in which the domain is divided into N equal meshes and the mesh size is Ax. u ax a22 (4.11 is given as.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. 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.12) and C 1 =N cL 1 e C2 ecL 1 .
16 is easily obtained from Eq.13) where the velocity u is positive. Equation 4.14 is the cell Reynolds number or cell Peclet number according to whether the diffusion constant a is the kinematic viscosity or thermal diffusivity.14) The parameter P in Eq. 4.54 Equation 4.15.15) Equation 4. (4. P(4). 4. where P =u a  l i24 + ii (4.14 may be recast to Eq.40) . 4.16) . i u x i+l2.15 and solved for in subsequent procedures. i+  i = (P+l)(4)  4)i (4. . i+l i = (P+l)( 1 . Equation 4.13 can also be solved with the given boundary conditions by reducing it to the following form.11 is finitely differenced using donor cell scheme for the convection term in the following.i + 4i2 Ax = a (4.
I_.(1+P) P (1+P) 0 + N .19) (1 . Both the analytical and numerical . 4.41 = N ) (1+P) [(1+P) N1  2) (4.19 into Eq.18) (l+P) + 1 Therefore. 4. as.1 (1+P) N 1 (4. (1+P) 1 . we obtain the solution for 4.16.17. i  0 = 4.17) 01 can be expressed in terms of 40 and 4N N N by Eq.12 and numerical solution Eq. 4.LI II . 1 (1+P) i .13. 4.'N1 (1+P) N1  ) 0 N . 4.~LI LIII~.~l*IIIIX1~~^ ~ ^I ^~Xi 55 Then.1 Inserting Eq.11 and Eq. (1 0 ) is ) P( N ¢0 (1+P) given as follows. 4.20) Now the analytical solution Eq. 2 1 = (1+P) (4 0) 4' N 92 = N2 (1+P) 2 (  0) = N. 4.20 have been obtained without any approximation for Eq.
i+l 2Ax il S S2ax 4) i+l 24i + i 2 i (4. 24 j. e (il)Pe (ePe1)2 NPe .20 is also substituted for 4.+1 i + ) il 0 (4. 4.22 we can obtain the effective Peclet number Pe of the finite difference solution in terms of the real Peclet number P as follows. in Eq. 4. 4.56 solutions will be reinserted to the finite difference equation.22) Comparing Eq.25) . 4. 2 ii P ( (1+P) (1+) N _ 1 N 4).24) It is also possible to go through the similar procedures with a central differencing form of the convection term in Eq. 4.13.23) (4. 4.21 and Eq.25.1 e Therefore. 4.13.13. Pe = in (1+P) (l+P) iIp2 N (1+P) . Eq. and the effective diffusion constant for the numerical solution will be obtained. 4.1 (4.25) (4. The in analytical solution Eq.12 is substituted for 4i Eq. + 24 S i i+1 i1 =e e l NP  1 (e P P 1) 2) ) ( (e 0 (421) The finite difference solution Eq.
26) where 2 < P < 2 The effective diffusion constant De can be readily obtained from the following relation.24 and Eq. 4. Therefore.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. 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 4.25 can be derived as follows.1. As in the previous section the velocity u is assumed to be positive and constant. it helps understanding the origin of the truncation error diffusion.27) In general. 4. Eq. 4. uLx D De Pe  (4. The convection terms in .57 The effective Peclet number for the finite difference solution of Eq. Although this approach is approximate and overpredicts the truncation error diffusion. 4.26 may be a good indication of the truncation error diffusion occurring in a general one dimensional problem. Pe = kn (2P) 2+P (4.
ae u( x where 0 <f < ae + u() ax i+  u~x a2 ) . 4. Pe also goes to zero resulting in no diffusion error in Eq.29 from Eq. If the Peclet number is small.29 hold only when the Peclet number is large. 4.30) Both Eq.26.e.29) i+ The truncation error diffusion can be quantified by subtracting Eq. 4.24 and Eq. the truncation error diffusion constant for a convection dominant problem may be given approximately as follows..( i+f 2 (4. the profile of 4 will be linear and the truncation error will be reduced to zero.28) Ax (4.4 and Eq. u i Ax i1 u i+ a4 u() ax u(a) ax (4. Therefore. i. 4.28 and Eq.31) ND . 4. 4. the convection is dominant over the diffusion.e.28.58 Eq. D uAX 2 (4. 4. 4. the diffusion is dominant over the convection. It can be shown that as P goes to zero.5 may be approximated as follows. i.
the total diffusion quantity of 4 due to the truncation error is not significant in comparison with that due to the crossflow diffusion. one is the truncation error diffusion and the other is the crossflow diffusion.33) . conservation equation with no source term is given in the following.59 4.3 Truncation Error Diffusion in a Two Dimensional Problem There are two sources of numerical diffusion in a two dimensional problem.1. Since the gradient of 4 in the flow direction is negligible in a convection dominant problem. The formulas for the truncation error and crossflow diffusion constants derived in this chapter reveal that they are approximatly of the same order of magnitude. 2 u + v 2 + +( ) (4. two dimensional. A steadystate.32 is finitely differenced using donor cell scheme for the convection term. u i 6x ilj + v i by ij_a(i+1 2 4x i + il1 +ij+l 2i 2 + 0iji by (4.32) = x 3y Equation 4. It will be shown that the truncation error diffusion in the direction normal to the flow cancels out and there is only a flow direction component left.
33.j+ + v(a) ly i+h. then the truncation error of the convection term is given as follows.2 is applied to Eq.y) is by rotation.n) ( = x cos6 + y sine I = x sine + y cose (4.x[cos + (cose tane 1 + sine) + sine tane ] ax xY (4. 4.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 = .v( ay ) ij .34) where tane = Ay/Ax tane1 = v/u The coordinate (x. 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 .1.35) .60 The approach in section 4. transformed to the coordinate ( .
2 2 2 + a = ( 2 ) (4. e is equal to 0 or 7/2. y v e Sx Then Eq. 4.32 is reduced to the following form. the problem is basically one dimensional and the truncation .36) where U = /u2v 2 Now Eq.35.61 where 5 is the coordinate in the flow direction.n) given by Eq.34 will be transformed to the coordinate system (t. dominant truncation error diffusion occurs When only in the flow direction.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.37) Equation 4. 2 (A)  [(sine tane1 + cose) 2 + a2 (sine + cosetan6 1 ) a + (0) 4 an U" (4. 4. 4.
The crossflow diffusion comes from the multidimensionality of the problem. In Fig.4. the hot and cold fluid enters the mesh from the left and bottom surfaces at 45* angle. It is entirely different Recently this addi from the truncation error diffusion. The hot fluid will come out of the top surface and the cold fluid out of the right surface. that will be named here as the crossflow diffusion. 4. In this section the origin of the crossflow diffusion will be illustrated and the corresponding diffusion constant will be quantified so that they can be used in the corrective scheme of section 4.62 error diffusion occurs in the flow direction as in a one dimensional problem. 4.2 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.l1(A). 4. which shows a single mesh with pure convection.3. tional false diffusion was explained by Patankar [52] and Stubley et al. [67. 4. 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.1. The origin of the crossflow diffusion is illustrated in Fig.1(A) is transformed into .68) and it was clarified that this is the dominant source of the error in most multidimensional problems.
1. 4. Illustration of the crossflow diffusion in a single mesh with pure convection .63 Hot Mledium Hot 7 Hot Cold (A) Cold (B) Cold (C) Fig.
4. diffusion occurs in multidimensional donor cell differencing of the convection term. 4. In the donor cell scheme it is distributed all .64 Fig. CAB which looks like a long onedimensional rod.&x u v ax Ay The effective diffusion constant for the crossflow diffusion will be calculated in the two dimensional Cartesian coordinate in Fig. ] = (lp)4 = where u . . Then homogeneous mixing occurs in the mesh. . in The numerical results Fig. of the value of 4 causes the crossflow diffusion.2 which shows a row of meshes aligned in the xdirection. This is a pure convection problem Consider the fluid element The fluid with no physical diffusion. A 3 . 4.1(B) are again interpreted as Fig. at those points is exactly equal to 4. and the sum of the values This distribution along the points A1 . . The value of 0 at the point A should reappear at the point X because there is no physical diffusion. 4.1(B) where the velocity components normal to the surface are considered. Another illustration of the crossflow diffusion is given in Fig.3. 2 (lp)[l+p+p + . 4.and intermediate temperature fluid will come out of the top and right surfaces. A 2 . an appreciable amount of false program user.1(C) by a Therefore. .
4. Illustration of the crossflow diffusion in a row of meshes aligned in the xdirection with pure convection .65 (1p)4' x p (1p) 2 p (lp) p3 (lp)4 p4 (1p) A5 Ay A Ax 2 4 y p t* U where p = Ax Ay Fig.2.
V A2 e A.3. Geometry for explanation of the crossflow diffusion in a two dimensional Cartesian coordinate . 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.66 4.
Diffusion A t1 . k2 = Ax sine/sin(el+8) k3 = Ax sine 1 /sin(e 1 +8) After substituting appropriate expression for each term. The diffusion occurs both to the The diffusion to the right hand side is considered first in the following. Gradient) A /k 1 =A of 4 Current 2 2U of 4 Therefore.4.t 0 PA £3 (Current) 2 1 constant ) where 1 = (Gradient) = U3 Lx/cose. p4A and (lp)4A and appears at the points Al and A 2 .67 element was on the line CAB at time t in the velocity direction. sine cose cos= sx Uacos61 sin(O+e ) I DRHSS . which shows that the value 4A diffuses out along the fluid element. In the donor cell scheme it is divided into two portions. right and left hand sides. The travel of the fluid element is analyzed in Fig. the following right hand side diffusion constant results. 4. v. and is translated The value 4A at the point A should reappear at the Point A'.
4.68 ti ..ct t2 t3 Fig. 4. Geometry for explanation of the crossflow diffusion in a rotated two dimensional Cartesian coordinate . Ya.
The crossflow diffusion constants are the ratios of the current and gradient of 4 in the x and y directions.40) .39) Equation 4. 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 D = uAx(lp) = VAy p (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. There are two diffusion components in a two dimensional problem as follows. 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 (4.38) Equation 4.38 is the effective diffusion constant when the diffusion is confined to the one dimensional fluid element.39 is identical to the following simple expressions.
4.5. Eq. 4.38 should be differentiated with respect to 6 in order to find out the velocity direction with maximum crossflow diffusion.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.6.41 shows that DRH S (or DLH S ) is maximum at an Tr7LHS angle of 61 < 6 e between 4 and e1. that is to say. (4. In the donor cell p' y scheme the value 4 is divided into three portions px 0 . When both 6 and 61 are equal to . 3 0 when tan 6 = tan6 1 .42. 4. < 6 < 81 or < . 4. Now the analysis will be extended to a three dimensional case in Fig.40 and Eq. 4.42 underestimates the crossflow diffusion as shown in Fig.70 where p = u Ax u v By Equation 4.42 give the same result.41) Equation 4. RHS D = It can be shown that. De Vahl Davis and Mallinson [24] also derived the effective crossflow diffusion constant and their result is given in Eq. D DDM = UAlxy sin2e= 3 3 4(by sin36 + Ax cos36) UAx tan6 1 sin26 3 4(tan8 sin6 3 + cos3) (4.42) Equation 4. In Fig. 4.
0 0 0 . 4.5.0t Dcf UAx de Vahl Davis & Mallinson Huh 81=8C 1.71 Cross flow Diffusion 2.0 150 300 450 600 750 900 Fig.5 80 \ % % 4 1.5 I I 0. 0. Comparison of the prediction formulas for crossflow diffusion constant by De Vahl Davis and Mallinson and Huh .
4.72 z C V Az X 0 o a X AY Ay B y x z C p . p Fig. Geometry for explanation of the crossflow diffusion in a three dimensional Cartesian coordinate .6.
t4 cos8. £4 cosy) nB . the following relation holds. and C instead of at the point X. XB.73 z occurring at the points A. point X is given as follows. n C denote The coordinate of the Any point on the plane ABC should satisfy the following Eq. B. This is the socalled crossflow diffusion phenoThe currents due to the crossflow diffusion are . menon.43) Since the point X is on the plane ABC.and nA unit vectors along those directions. X(£ 4cosa. along the directions XA. S= 4 + y z zAx 1 cos+ + cos Ay Ax + cosy Az (4. £1 = AX =/(x4cosa)2 + (4cos)2 4cos) 2 = Ax 2 + 42 2Ax£ 4 cosa . CX which are denoted as £i' z2' k3 can be obtained as follows. cosa cosB cosy (4. 4. BX. XC.45) The lengths of AX.43.44) 4 Therefore. S+ Ly Ax + Az = 1 plane ABC (4.
74 k2 = BX = /( 4cos)) 2 + (Ay_4cos) 2 + (k 4 cosy)2 =  y2 + + 42 L 4 £3 = CX = /( 4cosc) 2 + (£4cos8)2 '' + (AzL 4 cosy) / z + £42 2Azi + 4 cosy (4.48) The currents in the directions nA . nB. 2 o C 14cosy] = 1  £ 4 cosa.47) etc. nB' n C can simply be given as the products of the transported quantities and velocities.  n A = XA +1A  OA . I (Current)xAi XAI = px Px C1 1 Lt .OX (4. 4 cosy] 1 nB = [i4cosa. Therefore. z directions. n C can be obtained in terms of their components in x.46) + The unit vectors nA . y. (Az 4 cosy)] (4. (Ay4 cosS). Z 4 cos8.
/ Ax (Grad) = 4. Px D (Ax£ + Py D + nC(4 (Current)x 4 cosa) c osa) At ( 4 (cosa) (Grad)x = 4/Ax Therefore.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. (4. y. (Grad)x = 4.51) p Ax D x = lt (Axk 4 cosa) + p Ax ./ Ay 4/6x Q/by (Grad)z = 4/Azl (4. z directions are given as follows.75 I(Current) XB = Py p2 t I(Current) XCI where At = k /U z '3 At (4.(4 pz x co s a ) ct 4 + at ( 4 cosa) 4 .49) The gradients of 4 in the x.
54) 4.1 = uAx(lpx) (4. indices will also give the same result.53) The crossflow diffusion constants in the y and z directions can be obtained in the same way. This section is a review of the schemes that The have been used to eliminate the numerical diffusion. although none of them has been successful enough to be accepted widely. finite element method. we can simplify the result as follows.52) Inserting the expression for At given by At = £4 /U. Dy = vLy(lpy) Dz = wAz(lpz ) The rotation of (4. p Ax D = ux[ £4cos x .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. .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. most important ones may be the skew differencing. tensor viscosity method. etc.
The crossflow diffusion is not eliminated entirely in the skew differencing scheme. Basically the inclusion of four additional corner points contributes tomore accurate numerical modelling of the convection. In the skew differencing type schemes finite differencing of the convection term is done not in terms of the x and y directions. However.4 fall in the category of a corrective scheme.3. but just appreciably lower than that of the donor cell scheme. The corrective scheme is to reduce the diffusion constants in order to compensate for the additional false . Chang's method 1161 and Raithby's scheme 1533 are of a skew differencing type while the tensor viscosity method [25] and Huh's corrective scheme that will be introduced in section 4.1 can be categorized into two types.^ILli. S..II1~LI1I~~ ICII1_1___~1___*_____11__ 77 Most of the schemes in Table 4. In the skew differencing scheme the 5point relationship becomes a 9point relationship. skew differencing and corrective schemes... the interpolation between neighboring points may still give some crossflow diffusion. the interpolation should be done between neighboring points to obtain the value at the upstream point. Since the upstream point in the velocity direc tion does not necessarily fall on the mesh point where the quantity under consideration is defined... but directly along the velocity direction. The donor cell treatment of the convection term results in a 5point relationship in a two dimensional problem.~_ 1 i_.
thod of chaaceriics Fiite anuryic method Tenso.1.inte Conribo:r/me.ronLUD Lillinion/VUDCC Nuana.1QUICK t ETovh ci . 1981 [65) .e elements Clifft Dontcl t al. upwind f~iite di Teretn atl. Wilkes Elbahu ct l.ordc. £Ebzhau t aL/PATANKAR Lilinpion Orlandi Priddin Wada cit l.78 Chss' iction Cen:ered .thod LLxW'en doff  21 x11 80 x 400 Table 4.hod Mesh *txLnsformation rr. viscosity method Sel.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 !./LLUE Huint/QUICK Wilkes/LUE Lillimpon/SUD LiiL.. Esposito Glas and Rodi Huffenus Lnd Khali. List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research.Highe cdtr upwi:nd :"'ie Cdi't.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 .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 .ad.j. Grandotto Hickmott Firs.
The corrective scheme is valid only if the effective numerical diffusion constant can be predicted accurately. 4. Since the crossflow diffusion is dominant over the truncation error diffusion and the effective diffusion constants for the crossflow diffusion can be predicted theoretically. 4. where the balance equation is set up for a given control volume by considering convection and diffusion at the control surface. and D z in Eq. y and z directions will be reduced by the amounts of D . 4.79 diffusion. D. 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.53 and Eq.54.53 and Eq.1 Raithby's Scheme Raithby's scheme [53] is an Eulerian type of approach. The correction formula The diffusion con used here is Eq. the corrective scheme can give a numerical solution which is almost free from numerical diffusion. 4 = C1 + C2 n = C1 + C2 (  (4. The tensor viscosity method is one example of the corrective scheme. 4. stants in x.. 4.54.3.55) where n is the distance normal to the flow and U = /u 2 +v 2 . It is therefore a conservative scheme. although the validity of its correction formula is not clear.
4. A segment of. the calculation domain showing grid lines.80 Fig. control volume and notations in Raithby's scheme .7.
Although this scheme is nonconservative.56) where U = /u is the coordinate in the velocity direction and 2 +v 2 .2 Skew Differencing By Huh Skew differencing scheme here treats the convection term in a Lagrangian way. 'ij US (4.57) where the upstream value €US may be calculated by . not in terms of the convection that occurs at mesh interfaces. u 4 + 'v.3.81 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. The convection term is differ enced in the velocity direction between mesh points. it is much simpler than Raithby's scheme and becomes conservative in a unidirectional flow. which may be too complicated and time consuming for a practical purpose. 4. This should be repeated for the four control surfaces of every mesh in a two dimensional problem.= U ax dy SC (4. The inclusion of the four additional points also complicates the structure of the coefficient matrix in an implicit scheme.
58 for the numerical diffusion constant is apparently not consistent with the results in Eq. uv (4.3. the tensor viscosity [25) T = Lt uu u2 The expression for is given in the following.58) t2U= Eq.3.3 Tensor Viscosity Method The tensor viscosity method treats the numerical dif fusion constant as a tensor quantity. the crossflow diffusion constants D in section 4. 4. DD y . 4. 4.4 Corrective Scheme By Huh The corrective scheme is to use the reduced diffusion constants.3. DDx.53 and Eq. The validity of this scheme is based on the fact that the crossflow diffusion can be predicted accurately. D . 4.2.82 interpolating two neighboring points. D are derived . DD z in x. z directions to compenx sate for the crossflow diffusion effect. y.54. Some of the results of this skew differencing scheme is given in section 5.2. Dx = uAx(lpx) D Yy = Ay (lpy ) The expressions for . 4.
83 D z = wAz(lp z ) where px = u (4.+ w Ay Lz x etc. which is used with the cell dimensions to get the crossflow diffusion constants..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._IIU__lpULI~~I___~lLlt i~Lli ~i ~iL~pn~sl.3. . In a coarse mesh or recirculating flow field the latter implementation strategy is definitely preferred. The velocities at two boundary faces are averaged to get the representative velocity. they should be averaged again between neighboring mesh points to calculate the crossflow diffusion current at mesh interfaces.+ .Dz at the center of a mesh where all quantities are defined except the flow field. 4. After the crossflow diffusion constants are assigned to every mesh point. the mesh point implementation and mesh interface implementation.1 Mesh Point Implementation The mesh point implementation is to assign the crossflow diffusion constants Dx .4. Dy . Two implementation strategies have been tested so far.59) v u .
as shown in Fig. 4.8. 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. can be directly used with cell dimensions in Eq. The mesh interface implementation is always recommended although it has a more complex program logic than the mesh point implementation.Vl) and (U2. U and V. 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. 4.3. . The two velocity sets (Ul. For the second case the two outward velocities. The third case is expected to have a flow split as shown in the figure.V2) are used to give (DxlDyl) and(Dx 2 'Dy2 ). Dyl and Dy 2 . U1 and U2.84 4.2 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. The same logic can easily Appendix C includes be extended to a three dimensional case. There are three possible cases that should be treated independently in a two dimensional case. The right interface will have the crossflow diffusion constant (Dx1+Dx2 ) and the top and bottom interfaces.4.59. and the resulting D x and D y will be assigned to each interface. so that they are proportional to V1 and V2.
Three possible flow configurations for mesh interface implementation of the corrective scheme in a two dimensional problem ._~I~~ _~~~~_I__~__~~ __ 85 (1) (2) (3) Ul U2 V2 VI U U V14V2 VI+V2 V2 U2 = UV2 U = Ul + U2 TV2 Fig. 4.8.
60) Explicit (lD): n+l I n i At + u (i i x n i.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. 1. conservation physical constraint accuracy stability 2 ax 2 2 y 2 +ux +at + 7 + T7) + S sz (4. Explicit.86 the fortran program for calculating crossflow diffusion constants in an arbitrary flow field.62) . 4. 4. for the general conservation equation. Implicit and ADP schemes There are three typical solution schemes.4 Comparison of Explicit.) = i1 n D Ax ( i+1 2i l n n + n ) + iS i1 (4. 3.60. The pros and cons of the three schemes will be compared in the following four viewpoints. Implicit and ADI (Alternate Direction Implicit). 2. Eq. 4. which is a parabolic partial differential equation.
. 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. 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. + 4).63) 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 S2.1 Conservation The conservation principle is that the total amount of 4 should be conserved (S=0). when the net inflow or outflow at .
the same exchange term should be used for neighboring meshes. Although they usually give .4. there are some nonconservative solution schemes like ICE (Implicit Continuous Eulerian). where the convection terms of momentum equations are linearized in a nonconservative way. the conservative property is as important as the stability because the error from nonconservation may accumulate over many time steps. the conservative property may not be crucial. For a long time transient. implicit and ADI schemes can be made conservative by using appropriate differencing forms for the convection term. however.2 Physical Constraint The physical constraint is closely related with the stability in a practical sense. + CI 4. Although a conservative form is preferred to a nonconservative one. When we are concerned with a steady state solution. A scheme may be both stable and nonconservative and also both unstable and conservative because the conservative property is independent of the stability. All the explicit.88 the domain boundary is equal to zero. In order to guarantee conservation.
The laws of physics impose some restrictions on the numerical scheme to get physically reasonable solutions. 4. new values of 4 are given as follows. they are from entirely different origins.89 similar constraints on the time step size and mesh spacing. 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.1 Diffusion There is a maximum net flow that can occur between neighboring meshes by diffusion. )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 .2.4.D .
90 Since the relationship.64 or 4. 4. the solution will be unstable or stable with damping oscillations. . we have the following physical constraint. R AxAy(4) C) > 2D (R Ax CyAt 2 DAt X2 < 1  (4. Therefore the physical constraint should be respected in addition to the stability condition to get a meaningful transient solution.65) Ax Equation 4.64) Eq. 4. t' > Q' .64 or 4. Both of the oscillation and instaThe ADI bility should be avoided in a transient problem. d 2D t < 1 (4.65. however.65 may be a too conservative criterion in a real problem.63 is unconditionally stable.64 may be extended to a ddimensional case as follows. should still hold. 4. When the physical constraint is ignored and the time step size greater than that given by Eq. scheme given in Eq. 4.65 is used. there will be a damping oscillation in the solution when the time step size is greater than that given by Eq.
B 4y S The explicit and donor cell treatment of the convection term gives the convection quantity at interface S as ulAAtAy. Then uAttay is no more an appropriate expression for the convection quantity during At.2 Convection C A u B C A ). than the Courant limit.4.3 Accuracy Realistic interpretation of a finite difference equa tion comes from an integral form of the conservation . a portion of the control volume C will cross the interface S. It is based on the assumption that 4A is the repreIf At is greater sentative value of the control volume A. Therefore. 4. At < 'x u (4.4.91 4.2. the damping oscillation or instability may occur as in the case of diffusion.66) When a time step size greater than the Courant limit is used.
Profile assumptions of Q in space and time for various differencing schemes . ~~CIIL Is CIIII r r I I I I I! US Ds > I r I US DS Donor Cell Central Differencing Reverse of Donor Cell where US(Upstream) DS (Downstream) Fig.9. I I t i r _ I I [ n n+1 n n+l _ 0 It Explicit Implicit CrankNicholson $i+. 4.92 n+l n+l 4' I I.
whether the error introduced at a certain step will increase or decrease as the calculation goes over to the following steps. 0(Ax 2 ) is The also not appropriate for finitely large At and Ax. The accuracy expressed in terms of O(At). The Von Neumann . 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. The infinitesimal approximation of differential terms is not appropriate in most problems of our consideration. 4.93 equation. It depends on the maximum eigenvalue of the iteration matrix. In other words. However. appropriate schemes should be chosen according to those numbers or weighting should be done between neighboring time steps and mesh points to increase the accuracy of the solution. The stability can be checked by the theorems about matrix eigenvalues or Von Neumann analysis. If its absolute magnitude is less than one. accuracy depends on how well we can compute the convection and diffusion quantities at control surfaces over the time At. Since the exact profile depends on the Peclet and Strouhal numbers of the governing equation. the accuracy is usually the least important among the given four considerations.4.4 Stability The stability is a mathematical problem.
implicit and ADI schemes using the Von Neumann analysis. n+l i n €i+i I + u i+1 n ii 2Ax i_1 lt n n i+l . (4.67) The solution 4.94 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..68) .4.4. The stability is not a sufficient conIn the fol dition for an acceptable numerical scheme.4.2 2i + D i+1 i n 2 i i1 (4. 4.4. 4.1. stability criteria will be derived for the explicit. is expanded in Fourier series as follows. lowing sections. 1 4n = E~nei where I = .1 Explicit Scheme The Von Neumann stability analysis of a finite difference equation is simple in its idea.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.
67. in Eq.71) The function f(t) will be defined as follows. 4.70 may be rewritten as follows by defining cos6 x as t.69) Dbt Dt 2 Tx. 2 11 = [1 . Therefore.95 The absolute value of should be less than one for every When Eq. 4. the function .2dx(cosex) 2 + [Cxsinex 2 < 1 (4. [i . 4.68 is substituted results.72) In order to get the stability of Eq.70) Equation 4.2dx(t) where t = cose 1 < t < 1 ] 2 +C 2(1t 2) 1 (4. the following expression for = lCxsin xI where cx d x uAt A tx  2d (lcos x) (4.67. mode to achieve stability. f(t) = i 12 = (4d2_ 2 )t2 + (Sd 2+4d )t + C 2+4d 24d (4.
75) The parabola f(t) may have three distinct shapes according to the sign of the coefficient of the second order term. . 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. D(Discriminant) = (C 2 .2d )2 > 0 (4. f(1) = 0 (4. Therefore.10(1).1].73 can be solved by a graphical method using the characteristics of a quadratic equation. 4.1] Equation 4.73) for any t in [1.96 f(t) should be less than or equal to zero for any t in [1. f(t) < 0 (4. (1) 4d x 2 Cx 2 > 0 x : concave upward : linear 2 (2) 4d 2 C 2 = 0 (3) 4d x 2 Cx 2 < 0 : concave downward. The parabola f(t) always meets the taxis because the discriminant is always greater than or equal to zero.74) One of the two meeting points with the taxis is the point t=l because f(l) is equal to zero.
4.10. Three distinct shapes of the parabola.97 f(t) f (t) f (t) 'I I 4..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. f(t).. r I I t i t 1 1 t 0. according to the sign of the coefficient of second order term .
C (4. 4. .dx) is shown in Fig.1.2d Axis of symmetry: t = 4dx dx > C 2 x > . t=l as shown in Fig. 4.77) Summing up the results in Eq.4.78 is a finite difference equation with donor cell differencing of the convection term.77. 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.10(3). 4.4.11. we get the following stability conditions. 4.76 and Eq. 4d 2 C 2 4d C > 0 < 0 : : 0 < d dx > < 2 Cx The stable region in the coordinate system (Cx. 4d 2 .98 f(l) = 8dx(2dx1) < 0 0 < d < (4.2 One Dimensional Donor Cell Differencing of Convection Term The following Eq. 4. 4.76) 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.11.d ) for an explicit scheme with central differencing of convection term in a one dimensional problem . Stability ondition in the plane (Cx. 4.
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)
4. The region in the complex plane where (a +c ) should exist for the stability of an explicit conscheme with donor cell differencing of vection term in a general two dimensional problem .1 SRe 2 al +2 1+a +a 2 Fig. .104 Im \ .13.
.105 . .14.Re Both al and a 2 should be in the shaded circle to ensure that Ial+a 2 +1I< i. The region in the complex plane where the amplificatior factor. Fig. should exist for the stability of an explicit scheme with donor convection term in cell differencing of a general two dimensional problem .. 4.
the stable time step size for a given mesh spacing. and time step.88 and Eq. Therefore. where the points (Cx.88 and 4. 4. the stability However.e Rx . Ax.89 and al and a2 may be rewritten as follows. I 1 2 + (1+)e Rx Rx 2 + (1+)e a1 = Cx [1 + Ie 1 3 . At. the stable range of Cx should be determined in terms of the cell Reynolds number Cx/dx'. When Ax and At are already given. At.4 Stability Condition in Terms of Cell Reynolds Number The stability conditions in the previous three sections are in terms of the dimensionless numbers.d) should be in the shaded region for stability. which is independent of the time step size.e Y) (4. 4.89 can be solved by the graphical method as in the previous sections. The results are given in Fig. 4. The stability condition is again given by Eq.90) a 2 = C [1 y R R Y + R (4.4. which depend on both the mesh spacing. 4.16.91) . 4. 4. Cx and d . The stable region shrinks on a linear scale with dimensionality. our usual concern is can be checked directly.15.4.dx) and (C y.106 Both Eq.1. A three dimensional case can also be solved through the same procedures and the results are given in Fig.
4.idition in the planes (Cx.dx) and (Cy.dy) fcr an explicit scheme with donor cell differencing of convection term in a general two dimensional problem.15.107 d (d ) d = x 2 1 x + 1 4 C x(C ) 16 d= C2_ C x 2x x Fig. Stability cc. .
16.108 d xy(d or d z ) Cx d dx + 1 1 Cx (C d! S3 2 . Stability condition in the planes (Cx.dx) etc.(C Cx x y or Cz) z d ) Fig. for an explicit scheme with donor cell differconvection term in a general three encing of dimensional problem . 4.
88 and Eq. 4.94 ) D(Discriminant) = C 2 (K2C ) 2 > 0 (4.R and x t = cosex The stability condition is now that f(t) should be less than or equal to zero for any t in [1.C K(lt) + C ( l_1t = (C 22 KC 2 2 )t2 + (2C 2 22 22 K +C K)t + C 22 K  C K + C 2 < 0 (4. 4. and has distinct shapes for the Following three cases. [ . 4.C (+2x .) RX (1cose )) x + [C sine] x x < (4.1] f(l) = 0 (4.__LI___IIII__LI*X__ ID I 109 The following condition for Cx and Rx can be derived from Eq.92 can be simplified as follows.94 and Eq. f(t) = C (1+K) 2 (lt) .95 ) The function f(t) satisfies Eq.95.90. (1) Cx 2 K2 1) > 0 concave upward .93) ) where 2 K = 1 +.92) Equation 4. 4.
4. The same result will be obtained for the ydirection.17. V f .110 (2) C 2(K 21) = 0 linear (3) 22 C 2(K 2_1) < 0 concave downward For case (1) and (2). therefore the subscript x may be replaced with the subscript y in Fig. to zero. solution procedure is similar to those in the previous sections and the results are given in the following as.97) The results in Eq. The stability condition in Fig. 4.97 are shown in Fig.yu < 1 (4. R > 0 x : 0 < C < x 2 2 2(1+) x and (4.96) R 1 x < 0 < C x < (1+2) Rx . (4.17 is applicable to all general 2D problems. t=l.17. 4. although it can be relaxed if there is a constraint on the velocity direction and mesh spacings such that the following' relation is satisfied.98) Ax . 4.96 and Eq. f(l) should be less than or equal For case (3) the axis of symmetry of the parabola The should be on the right hand side of the point. 4.
17.Cy) for an explicit scheme with donor cell ditferencing of convection term in a general two dimensional problem .111 x (C y C x  2 1 2 (1+ L) R x C = 2(1+ ) x 2 0 R (R x y Fig. Stability condition in the plane (Rx.Cx) and (Ry. 4.
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/Lx . 4.18.112 1 1 x CX x 1 (l+f) (1+ x ) 1+f 1 1 f 0 x V where f =y u Ax Fig.
17 and numerical experiments.20. n+l n t n+l n+l n+l n+l n+l S i + u i +xu i i= D i+l D i Lx 2 ii (4. 4. 4.2 shows the com 1 l+f in Fig. 4. parison between the analytical result in Fig.17 should be Table 4. 4.99) The stability condition is shown in Fig. 4.113 Then the number replaced with on the C axis in Fig. 4.18.2 Implicit Scheme The following Eq.uLx. d >  greater Cx (4.19 and Fig. the stability condition is found to be the following.Cx d > . and it can be seen that the implicit scheme is unconditionally stable if the diffusion constant is than . 4.C Y y . (4.101) (4.4.99) (4.102) dx > .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.
083 Von Neumann Analysi s 0.499 0.488 0.333 4.147 0.833 8.720 0.019 Numerical Experiments 0.660 0.2.780 0.120 0.3 8.540 0.403 0.333 83.260 0. R x 0.660 0.180 0.060 Table 4. 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 .114 Maximum Stable C.629 2.380 0.120 0.333 833.167 2.
dx) for an implicit scheme with donor cell differencing of convection term in a one dimensional problem . Stability cndition in the plane (Cx.115 d =  x d =.19.C x 2x x 2x C x dx 1(C2+C 2xx ) Fig. 4.
4.20.116 Cx=1  2 Rx 2 1 0 R x R C x R +2 Fig.Cx) for an implicit scheme with donor cell differencing convection term in a one dimensional of problem . Stability condition in the plane (Rx.
The Von Neumann analysis of the first formulation in Eq.4. One of the formulations has both the x and y direction components in each step. 4.+d )] C ) {[1+(lcosex +dx)) +[fsinex }{[l+(1cosy) (+dy) + Y[siney } (4.63.4. 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. the following condition should be satisfied. 4.103) jsj <1 In order for I.63 gives the following stability condition.117 4. C d x>2 (4.104) .jto be less than or equal to one.3 ADI Scheme There are two possible formulations of the Alternate Direction Implicit (ADI) scheme given in Eq. 2 _C C 2 {[(cos )(+d X))] C [sine I2} {(1cos6 C 2 )(.
. C d y>  2  4 (4. 4.63 has the same stability condition as the fully implicit scheme. C d x > 2 2x (4.118 The same condition should hold for the ydirection.63 can also be given by the Von Neumann analysis as follows. 4.107) Therefore the second formulation of the ADI scheme in Eq.105) The stability condition of the second formulation in Eq.106) C d y  >  ~ 2 (4.
1 Updating of the Reference State The state of air is determined by any two independent properties.1 Natural Convection Results The containment is modelled as a two dimensional rectangular compartment with an obstacle as a heat sink. 5. 5. density is determined by temperaSince the natural convection of air ture and pressure. 5. occurs as a result of the buoyancy force due to density .119 CHAPTER 5 RESULTS Some calculations are performed to validate the physical models and numerical schemes in the previous chapters. Natural convection occurs due to the heat transfer between the obstacle and containment air. energy convection and heat transfer models are considered and the resulting flow fields are given in Fig.1 and Fig. For example. Finally. 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.1. 5.2. Updating of the reference state. the ADI scheme is explored to use a time step size larger than the Courant limit to save the overall computational time.
= 677.4 sec AT = 100F T .29 Ibm A= 2 ft2 T.0 Btu/lbm 0F (Q ob = 208. = 680F air T = 580F Fig.5 ft/sec h = 1.120 p ).1.8 sec . Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 10.8537x10tu/ft20 F sec (Cp) = 1. 5. 0.
Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 62. T ab = 580 F Fig.0 Btu/lbm OF (c )b = 208. = 677. 5 ft/sec h = 1.4 sec 1 0 AT = 10 F T air = 680F .121 > 0.8537x10 tu/ft2 F sec p)ob = 1.29 ibm (K 2 A= 2 ft T.5 sec .2. 5.
total Pold Told new old Qcell cv Mf.1) T Pnew = old + Ttotal c v M f.122 gradients. The density change of liquid.cell where Pold' Pnew: Density of the fluid at old and new time steps. however. new Pnew RT new (5. c : Specific heat of the fluid at constant volume. time Pressure of the fluid at old and new steps. temperature and pressure are equally important parameters in determining the flow field. Pold' Pnew: . not on the pressure because of a high modulus of elasticity. depends primarily on the temperature. Told' Tnew: Temperature of the fluid at old and new time steps. The equation of state of an ideal gas is given in the following.
but they may also be based on the vertical height of the obstacle.1. The Nussett and Grashof numbers here are based on the axial length of one computational mesh. the atomic gases have different values of k..1. Mf. 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. 5. .cell. An experimental correlation is given in Eq.4 for air.2 in terms of the dimensionless numbers.3 Heat Transfer Modelling The heat transfer to the obstacle is modelled as a natural convection from a vertical surface. 5.2 Energy Convection Another difference between liquid and gas is the ratio of the Specific heats. . 5.total : Mass of the fluid in one cell and in the total domain.0 Monatomic. The value of k is 1. diatomic and polyTherefore. Qcell total : Heat input to the fluid in one cell and in the total domain. The energy convection will be underestimated by a factor of 1.123 Mf. convection energy transfer should be in terms of enthalpy instead of internal energy. k = c /C for liquid and 1.4 when internal energy is used for enthalpy in air flow.
555 0.2179 x 10 B = 1.2) 3 2 Gr = gS(AT)L p 2 (5.sec . Gr = 9.0763 ibm/ft 3 g = 32.021 n 0.124 Nu = C(GrPr) n ( 5.923 x 103 Therefore. h = 1. This may be a typical value for the natural convection heat transfer in the containment without any phase change.876 The heat transfer coefficient is given as.4 lbm/ftsec Btu/ft 2 F.4 The following data are used to calculate the heat transfer coefficient at the obstacle wall.8537 x 10 .109 > 109 C 0. .2 ft/sec' 5 V = 1.25 0. L = 2 ft AT = 500 F p = 0.7134 x 108 Nu = 89.3) Pr = 0.708 (air) GrPr 105 .
pf Pob: c h : Fluid density.4 Turbulence Modelling Laminar flow is assumed in obtaining the results in Fig.4) where CQ is the source term in the energy equation of fluid.1 and Fig. CQ = Cpob (TT) (5. Specific heat at constant pressure. can be obtained from heat balance between the obstacle and fluid as follows.5) where M A : Mass of one computational mesh.1. Therefore. Heat transfer area between the obstacle and fluid over one computational mesh. Heat transfer coefficient between the fluid and obstacle.4 sec. 5. the heat . 5.2. If the flow is turbulent.125 The VARR input [21) requires the time constant of the heat transfer instead of the heat transfer coefficient. (Mc )ob (MC) obf hApo b ob f (5. T = 677. Obstacle density. 5. The time constant. T.
The results in Fig. 5.3(A) is parallel with the mesh orientation so that no crossflow diffusion occurs. is compared with De Vahl Davis and Mallinson's in predicting the magnitude of crossflow diffusion. 5. Figure 5. The skew differencing and corrective Huh's formula schemes are tested for various problems.1 Truncation Error and Crossflow Diffusion The truncation error diffusion is compared with the crossflow diffusion in a two dimensional.7 show that excessive numerical diffusion has occurred due to the .2. 5. 5.6 and Fig. 5.4 and Fig. 5.3(B) is skewed at 450 to the mesh orientation and both truncation error and crossflow diffusions are included in the solution error. The flow in Fig.3. The flow in Fig.126 transfer.to the obstacle will be enhanced and the diffusion transport of momentum and energy will also be enhanced in the containment air. 5. The corrective scheme is tested in recirculating flow problems for mesh point and mesh interface implementations. steadystate problem in Fig. 5.5 show that the numerical solutions is close to the analytical solution and the truncation error can be neglected.2 Numerical Diffusion The truncation and crossflow diffusion errors are compared in a simple geometry where an analytical solution can be obtained.
0 B T specified Fig. 0 ST By =0.127 (A) .0 u = v = 7.0 BT S=0. (A) parallel flow and (B) cross flow.0 ay AB=BC=5.0 B T_ 0.0 x BT  @x a( aT + 2 2 a2 ay ) Th= 10.3.0 u = 10.0. Problem geometry for the two cases. for evaluation of numerical diffusion . 5.0 (A) (B) AB=BC=5.07 (B) =0.0 a= 1.
5.128 Analytic solution Numerical solution with Donor Cell T 10. Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 10x10 meshes along the line CD in Fig. spatial coordinate Fig.4.3(A) . 5.
__LL l. 5.^XI ill I~LX_L~LI.3(A) .~ I~.5.__~ .~l r. Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CD in Fig. 5.~ 129 Analytic solution Numerical solution with Donor Cell C spatial coordinate Fig.
5. Comparison of the analytic solution. 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. 5.6.3(B) .130 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Df cf T 10 spatial A coordinate Fig.
. 5.J 0 I' C spatial A coordinate Fig. analytic solution with increased diffusion constant and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CA in Fig.3(B) s' / line CA in ./ Fig.3(B) . Comparison cT the analytic solution..7.131 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Dcf T 10 / I / / . 1/ 5. 5.
Figure 5. 10 x 10 meshes and Fig. it is necessary to test them for various cases.132 crossflow diffusion.6 are for 5. 4. 5.54. It can be seen that the former result is much better than the latter. 5. 5.2 Validation of Huh's Correction Formula The crossflow diffusion can be predicted by Huh's formula. Figure 5. solution. 5.2. 5.5 and Fig. The problem geometry in Fig.53 and Eq. Since the corrective scheme is based on the validity of those correction formulas.11 and Fig.4 and Fig.9 is the result by Huh's formula and Fig.10 by De Vahl Davis and Mallinson's for the problem in Fig.8 has the arbitrary angles of 6 and 61 such that 6 = 600 and 61 76.6 and Fig.12 for a simple pure . De Vahl Davis and Mallinson's and Huh's formulas give the same crossflow diffusion constants for the case. 5. 6 = 6 = 450.7 than in Fig. 5. meshes. 4.810.8. 5.3 Skew Differencing Scheme Two skew differencing schemes. 5.2. 5.7 are for 6 x 6 It is seen that a finer mesh gives a better Since the crossflow diffusion constant is pro portional to the mesh spacing. Raithby's and Huh's.6.7 show that the numerical solution can be reproduced by the analytical solution with increased diffusion constants by the amount given by Huh's formula. Eq. more diffusion has occurred in Fig. are tested in Fig. 5. Figure 5. 5.
Problem geomctry with arbitrary angles of 6 and 81 to compare De Vahl Davis and Mallinson's and Huh's formulas for prediction of crossflow diffusion constant .2406 by = 1. 5.0 Ax = 0.0 v = 8.6603 DDM= 0.335 Huh A e1 = 76.7674 DH u h = 3.8.133 Yo 2 T=0 D a = 1.810 8 = 600 8 Meshes 24 Meshes Fig.0267 u = 5.
(D+Dcf).8 . by Huh's formula along the line BD in Fig. 5.134 Analytic solution with D+Dcf Numerical solution with Donor Cell T 10. 5 0 spatial coordinate Fig.9. Comparision of the numerical solution with donor convection term and cell differencing of analytic solution with increased diffusion constant. 5.
by De Vahl Davis and Mallinson's formula along the line BD in Fig.10. Comparision of the numerical solution donor cell convection term and analytic differencing of with increased diffusion constant. solution (D+DDM).135 Analytic solution with D+DDM Numerical solution with Donor Cell T 101 spatial coordinate Fig.8 . 5. 5.
75 81.25 65.75 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 34.5 0 6.11.63 87. 0 100 100 100 100 100 50 100 50 0 50 0 0 V Fig. 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 convection problem with e = 45* .25 18.25 0 Skew differencing by Huh (True solution) 100 100 100 100 50 0 0 0.136 Donor cell 100 100 100 100 100 93.38 50 31.75 25 U 12. 5.5 68.
91 71.3E 21.12.26 80.86 16.A Skew differencing by Raithby 100 100 100 100 96.137 Donor cell 100 100 100 100 50 .43 7.25 0 0 0 .5E 25.70 4.56 1.52 1. 5.85 0.430 0 50 0 Comparison of the true solution and solutions by donor cell scheme.67 5.9 33.1 6.83 72.25 37.25 53.11 0 11.8922.6.33 70.430 .3: 11.13 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 6.96 17.99 10.6 0 Skew differencing by Huh 100 90.59.74 38.50 2. 62.11 3.25 1.3' 40.22 9.91 31.88 81.01 55.71 20.2! 90.47 23.5 0 0 63. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure covection problem with e = 63.E 21 0 12 True solution 3.5C 100 100 50 Fig.
138 convection problem. however. Fig. 5.3. the However. 5.14 show that the analytical solution can be reproduced by Huh's skew differencing scheme for the case. both of them are conservative in a unidirectional flow.1 and 4.2.13 and Fig. 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. 5. method while Huh's scheme is Lagrangian. The purpose of the corrective scheme is. former is conservative while the latter is not.3.12 has a flow at a 63. Raithby's scheme is an Eulerian Therefore.430 angle and neither Raithby's nor Huh's scheme can reproduce the true solution. It is indicated that Raithby's scheme may give unphysical results in problems with steep gradients of the quantity under consideration. 8 = 1e= 45*.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. but less than inthe donor cell scheme.2. Some crossflow diffusion has occurred in Huh's scheme.11 has a flow given in sections 4. The details of those schemes are Figure 5. Figure 5. to obtain the true solution by subtracting the additional crossflow diffusion .
13.139 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh C A spatial coordinate Fig. 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.
140 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh 10 5 0 spatial A coordinate Fig.14.3(B) . 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 CA in Fig.
141 constant from the total diffusion constant of the finite difference equation.19 is a hypothetical 3 x 3 recirculating flow field where the inlet boundary values are specified on the left and bottom surfaces. Two implementation strategies of the corrective scheme. The corrective scheme is also tested in the problem geometry of Fig. Figure 5. 5. 5.20 shows the diffusion corrections of the mesh point and mesh interface implementations at each interface.15 numerical solutions are given for a simple pure convection problem with various diffusion corrections. Figure. The corrective scheme gives a physically reasonable solution. Fig.17 and Fig.3(B). In Fig. The flow is at a 45* angle with respect to the mesh orientation.21 gives the true solution. The full correction of the crossflow diffusion gives the true solution without any numerical diffusion error.430. mesh point and mesh interface. 5. although not identical to the true solution. donor cell solution and the solutions by mesh point aand mesh interface implementations. The mesh point implementation gives an unphysical solution .16 has a flow with 8 = 63. are introduced in Chapter 4. 5.18 show that the crossflow diffusion can be eliminated by the corrective scheme and that a finer mesh spacing always gives a better solution. Both of them are tested in a recirculating flow problem with pure convection. Figure 5. 5. 5. Fig.
0 10 7.08 5.13 7.0 1.88 10 5 0 2.71 0.5 x y Fig.92 5.15.1 10 8.3 D=0.0 2.25 0 5.29 5.59 7.0 2.12 5.5  3.88 5 4.0 8.29 0.75 6.99 D=0.41 u=v=l Ax=Ay=l D =D =0.5 (True Solution) 9. Donor cell solutions with various diffusion corrections for a pure convection problem with e = 450 . 0 V ________ D=0.0 2.142 D=0.71 5. 5.5 0 1.01 7.
11 0.56  2. True solution and donor cell solutions with and without diffusion correction for a pure convection problem with e = 63.143 D=0.16.33 1.11 True Solution V 63.37 3.93 10 5.59 1. v=2 x= y=l 2 D =D = x y 3 Fig.430 u=l.04 4.73 10 7.41 0.07 2.07 10 3.11 6.82 0.91 0.07 0.10 9.430 . 0 D2 3 0.96 3. 5.
3(B) . Comparison of the analytic solution. numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig.144 Analytic solution Numerical solution with donor cell Solution by corrective scheme 10 A spatial coordinate Fig.17. 5. 5.
numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig.145 Analytic solution Numerical solution with donor Solution by corrective scheme 10 5 0 spatial coordinate Fig. 5.3(B) .18. 5. 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.
5 0.75 1.625 0.2 0.75 0. 5.20.19 .25 1.625 0 .75 0 0 0.75 1. Diffusion coastant corrections at each interface for the mesh point and mesh interface implementations of the corrective for the recirculating flow field given in Fig.75 1.147 0.125 0. 5.2 0 0 0 mesh interface implementation Fig.25 mesh point implementation 0.5 0 0 5 0 0 .
51 10 5 0 Fig.16 11.10 9.21 7. 5.21 mesh interface implementation 0. 5.148 True solution Donor cell mesh point implementation 5.75 6.82 8.26 8.94 9.83 6. 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.10 8. Comparison of the true solution.08 9.86 12.19 .21.18 7.16 6.
23. respectively.2 kg/sec. 5. The steam concentration distribution at the beginning is given in Fig. 5. The result in Fig.27 and Fig.25 and Fig. 5. 5. 5. 5.22 shows the flow field in the containment and more detailed information about the flow field is given in Fig. Figure 5.25 and Fig.19. 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. flow field is used here to test the mesh point and mesh interface implementations of the corrective scheme. Since the steam concentration distribution and flow field are given at the start of the calculation.149 while the mesh interface implementation gives a physically reasonable solution.3 sec with the corrections in Fig. 5. rectangular containment with an arbitrary recirculating flow field.26. The true solution cannot be reproduced because some of the information about the flow field has already been lost in Fig. 5. the steam concentration for the new time step can be calculated in an explicit way.26 show the diffusion constant corrections of the mesh point and mesh interface implementations and Fig. 5.28 is not a reasonable solution while the mesh interface implementation result in Fig.27 is physically reasonable and even better . After a while a flow field is set up in the This containment by the input mass and momentum of steam.24 in which the source mesh has the highest steam concentration. Another test calculation is done for a two dimensional. 0. 5. Figure 5.
0m Fig.150 Ax = 0.5m Ay = 1. 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.
2214 velocity [m/sec] Ax = 0.7126 0. 5955 SOURCE 0.0492 0 .4920 0 .1139 0 .34 0.5174 1. 0648 1. 045 0. 0735 0.2216 0.531 0.4870 0.1229 0.0937 0 1.2159 0.1276 0 .3074 0.2474 04 1379 0. 1317 0. 5.3253 0.151 0.1422 2148 0.2435 0. Detailed flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .23.3 095 1.2460 0.456 0.1385 0.207 2.2629 0.0348 0.0127 0.93 2.2190 0.5893 0. 861 1.5 m Ay = 1 m Fig.3957 0. 3563 0.
1085 0.24.0298 0.0620 0.0049 0.1750 0. 5.1266 0.0136 0.0346 0.1710 0.0095 0.0676 0.2 kg/sec of steam Fig.0358 0.0819 0.0179 0.0007 0.0031 Ax = 0.1712 0.0002 0.0305 0.152 0.0011 0.1420 0. Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .1859 SOURCE 0.1244 0.1592 0.1178 0.5m by = Im Source: steam concentration in Ikg/m 3 0.
Diffusion co.0348 0.0311 0.0 278 0.C209 0. 0.1178 0.0937 0.0193 0. 0086 0.0794 0.1452 0..2740 0.1701 0.0 362 0.1014 258 0.1451 0.1279 0.stant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .1583 0.0590 0.0311 0. 5.0440 0.5m by = Im Diffusion constant in Im2/sec] Fig.0253 0. 0396 0. 0152 0. 0488 0. 0. 0227 0.0255 SOURCE 0.1763 0.0224 0.0600 0. 2422 0.0 183 Ax = 0.3262 0.0731 116 0.1854 0.1 389 0.1694 3106 0.25.153 0.0300 0. 0.
0479 0.0287 0.0632 0.0 111 0 Lx = 0.2529 0.1149 0 0 0.1026 0 SOURCE 0 0 0.0 257 0.2134 0.5m by = lm Diffusion constant in Im2/sec] Fig.4732 0.154 0 0 0.0860 0.2317 0 0. 5.0 76 0.0518 0 0 0.0443 0.0534 0. 0706 0. Diffusion constant corrections of the mesh interface implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .0215 0.3694 0.0192 0 0.0 923 0 0.26.9769 0 0 0. 0579 0. 0240 0.
0947 0.0342 0. 0.1135 0.0003 0.1363 0. with donor cell differencing convection term without any diffusion of correction .1321 0.0046 Lx = 0. 5.0833 0.1243 0.1623 0.0233 0.0011 0.1733 0.0735 0.3 sec.0107 0.27.2 kg/sec of steam Fig.1711 0.0195 0.0448 0.0016 0.155 0.1475 0.0465 0.1706 0.5m Ly = Im steam concentration in [kg/m 3 ] Source: o.0068 0.0342 0. Steam concentration distribution after one time step.1847 SOURCE 0.
1797 0.0433 0.1911 0.156 0.0005 0.1729 0.0042 Lx = 0.0709 0.2086 SOURCE 0.28.3 sec.0250 0.1247 0.1240 0.1747 0.0029 0.1646 0.0287 0.0815 0.0010 0.1493 0.0355 0.0229 0.0092 0. 5. with mesh point implementation of the corrective scheme .5m Ly = Im steam concentration in Ikg/m 0.0068 0.0465 0.0937 0.1334 0. Steam concentration distribution after one time step. 0.2 kg/sec of steam Source: 3 Fig.
0424 0. 5.1756 0.0007 0.0039 ax = 0.0192 0.3 sec.29.2 kg/sec of steam Fig.0010 0.1342 0.1243 0.5m by = Im steam concentration in [kg/m 3 Source: 0.1310 0. Steam concercration distribution after one time step. 0.1112 0.1794 0.0295 0.0100 0.0815 0.157 0.1628 0.1738 0.0947 0. with mesh interface implementation of the corrective scheme .0481 0.0058 0.0722 0.0003 0.0228 0.1847 SOURCE 0.1487 0.0360 0.
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. 5. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable.3 ADI Solution The ADI scheme is tested to use a time step size larger than the Courant limit. The results of the ADI and explicit schemes are compared in the problem geometry of Fig. when the Courant limit is 1 sec.32 shows the profiles of 4 at the same loca tion for the time step size of 1 sec.5 sec. 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. In the meshes at the top of the source mesh. it has the maximum time step size that can give a physically reasonable solution.30.0. which is exactly the .0 and the inlet boundary value of 4 is. 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. Figure 5. In Fig. The ADI and explicit schemes give almost identical results during all the transient. 5. Figure 5.31 shows the profiles of 4 along the dotted line for the time step size of 0. 5.158 than the donor cell solution. =10.30 there is an inflow from the bottom surface at the velocity of v=1.
0 Fig.0 c= 1. 5.1I I _~ 1_. v = 1.0 Ax = Ly = 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 on this line are compared.30.YXI~^.
0 11.0 Fig.0 12.160 12.0 11. 5.31.0 10.0 11.5 .0 A iatlai B 10. Comparison of the explicit and ADI solutions for along the line AB in Fig.01 12.0 coordinate 11.0 10.0 11.5 sec 12.0 11.0 sec for Lt=0.0A 12.30 the profile of sec and the Courant limit of 1.0 12. 0A Explicit ADI 10.0 0.0 10. 5.0 10.
0 11.0 10.0 12.0 10.0 4 sec 11. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig.0 10.0 . 11.0 sec and the Courant limit of 1.0 Explicit ADI 10.0 10.0 11.0 16 sec 20 sec 12.161 1 sec 12.0 11.0 ~ spatial coordinate 12.0 12.0 Fig. 5.30 for At=1. 12.0 sec .32. 5.
33. 5. 5.31. there is some deviation between the two results. Initially. 5. 5.33 shows that the ADI scheme can use a time step size up to five times the Courant limit. 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.32 and Fig.162 the Courant limit. . but it damps out rapidly and the ADI scheme gives almost identical results with the explicit scheme. Fig. No crossflow diffusion corrections are involved in the results of Fig. Fig.
163 12.30 with different time step sizes .0 At =0.0 sec along the line AB in Fig. Comparison of the ADI solutions for the profile of at t=10.5 sec ec 11.33. 5. 5.0 t spatial coordinate B Fig.
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. in the finite difference donor cell treatment of convection. The mass diffusion. and turbulence equations are solved separately using that flow field. energy.164 CHAPTER 6 CONCLUSION 6. The con tinuity/momentum equations are decoupled from the scalar transport equations and solved by the SMAC scheme to obtain the flow field. The . The total diffusion con stant is the sum of the laminar and turbulent diffusion constants. 6. Crossflow diffusion occurs due to the donor cell treatment of the convection term in a multidimensional problem. or superheated steam as may be appropriate. truncation error and crossflow diffusion.2 Numerical Schemes (1) There are two numerical diffusion sources. The thermal equilibrium is assumed and the phase change occurs to maintain 100% relative humidity.
have been tried in order to eliminate the numerical diffusion. The skew differencing scheme gives good results for some problems. The truncation error diffusion occurs in the flow direction while the crossflow diffusion occurs in the diagonal direction of neighboring mesh points. but not always. Therefore. . (2) Two types of approaches. the corrective scheme is generally preferred to the skew differencing scheme. recirculating flow problem is due to the crossflow diffusion.165 effective diffusion constants of the truncation error and crossflow diffusion are of the same order of magnitude. It is conservative and an explicit scheme can be used without affecting the simple solution structure. Therefore. The gradient of the scalar quantity under consideration is usually small in the flow direction in comparison with that in the direction normal to the flow. The inclusion of corner points complicates the matrix structure and a fully implicit scheme should be used for maintenance of stability. The corrective scheme is based on the fact that the additional crossflow diffusion can be predicted theoretically for every mesh at every time step. most of the numerical diffusion error in a multidimensional. ~unx. convection dominant. It gives unphysical results for most recirculating flow and coarse mesh problems. skew differencing and corrective schemes.
First.166 (3) Two implementation strategies. i.4. 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.e. (5) The maximum time step size is limited by the Courant The Alternate Direction condition in an explicit scheme. The mesh interface implementation is always preferred to the mesh point implementation in a recirculating flow problem. The graphical method is used to obtain the stability condition in terms of the Courant. it has its own limitations due to the following factors. the physical constraint in section 4. (4) The Von Neumann stability analysis is applied to various finite difference forms of a general conservation equation. The results of numerical experiments are found to be consistent with the Von Neumann analysis. mesh point and They mesh interface. Implicit (ADI) scheme can be used to increase the time step size and decrease the overall computational effort.. Cx. using the characteristics of a quadratic equation. are identical in a unidirectional flow. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable. diffusion and cell Reynolds numbers. dx and Rx .2 imposes a maximum time step size that can give a physically . are tried for the corrective scheme.
asymmetry of the given problem may give different results according to the sweeping sequence. Second. .167 reasonable solution. the ADI scheme is a fractional step method. a steadyThe state solution may never be reached due to the fact that one time step is composed of a few fractional steps. For example.
168 CHAPTER 7 RECOMMENDATIONS FOR FUTURE WORK 7.g. time step size. transitional and laminar flow regimes. 7.2 Numerical Schemes There have been two approaches. etc.. multidimensional recirculating flow problem. A criterion for the flow regime should be developed in such a form that it may be implemented into a computer code.1 Physical Models Since both the laminar and turbulent flow regimes are possible in a complicated geometry and flow field. to eliminate the crossflow diffusion. The flow regime criterion may be replaced with a model that covers all the turbulent. The modified kE model in a low Reynolds number flow may be a guide in this approach. . velocity gradient. distance from the wall). The model should show proper limiting behaviors as the Reynolds number goes to infinity (turbulent) and zero (laminar). it is necessary to predict the flow regime in order to use appropriate physical models. skewdifferencing and corrective schemes. The important parameters may be the velocity. geometry (e. They may still be improved further to get an accurate solution in a transient. previous time step information.
2. More tests and validation calculations are required in recirculating flow problems. ADI and implicit schemes although only the implicit scheme has been used up to now. The solution in Appendix A may be a useful guide in this work.169 Corrective scheme The corrective scheme has been successful by the mesh interface implementation strategy in the scope of this work. . It may be possible to develop a reasonably simple form that may or may not include the corner points. The stability of the skewdifferencing scheme is open to question in the explicit. Skew differencing scheme 1. The conservative form of the skew differencing scheme is complicated and time consuming in comparison with the corrective scheme.
F.User's Manual. "A Finite Difference Scheme for the Incompressible AdvectionDiffusion Equation. W. J. and McDonald.H.A.H. "Transport of Turbulence in Numerical Fluid Dynamics. Briggs. Broadus. E.R. Briley. John Wiley & Sons (1960). 94110 Amsden. "The SMAC method: A numerical technique for calculating incompressible flows... "Hydrogen Distribution in a Containment With a HighVelocity HydrogenSteam Source.R. D. 393409 (1980).. "A New Finite Element Formulation for ConvectionDiffusion Problems. et al. B." Journal of Comp. Bloom." Comp. Methods in Applied Mech.R... Engineering Calculation Methods for Turbulent Flow.. "Solution of the Multidimensional Compressible NavierStokes Equations by a Generalized Implicit Method. Phys.R. G.." MIT Ph. and Whitelaw. and Patankar." LA4370 (May 1970). Phys. A. and Eng. and Lightfoot. Bird. Bradshaw. A. H. Phys. "Transient Effects in Turbulence Modelling. Briley.R. 5473 (1980). Stewart." Numerical Heat Transfer 3. F. "BEACON/MOD 3: A Computer Program for Thermalhydraulic Analysis of Nuclear Reactor Containments . and Harlow. and Harlow. Comp. 34.170 REFERENCES [1] Amsden.R.. 6.." HEDLSA2682 (October 1982).H. R.B.D.V.. et al.. 332325 (1970).. A. Amsden. H. W. F. Cebeci. and Harlow.. and McDonald." Journal of Comp. Academic Press (1981). Boyle. C. W. 372397 (1977). 3.N.A.G. S. thesis (1979). 233241 (1975). (April 1980). T.A. Phys.. 24. 6. [2] [3) [4] [5) [6] [7] [8) [9) [10) [11) [12] .H. "On the Structure and Use of Linearized Block Implicit Schemes." Journal of Comp. P. D. "A Simplified MAC Technique for Incompressible Fluid Flow Calculations.E." NUREG/CR1148 (EGG208). Transport Phenomena. Baliga." Journal of (1968).
H... "Studies of Numerical Methods for the Plane NavierStokes Equations.I. 32. 2943 (1976). thesis (1981).A Computer Program for Calculating Timedependent Turbulent Fluid Flows with Slight Density Variation. De Vahl Davis. Chang.S. P. Chen. S..171 [13) Carbajo.D. and Mallison. Cheng. "SecondOrder Numerical Advection. Methods in Applied Mech.B. Cloutman. Cook.. G." MIT M. "Comparative Analysis of Numerical Methods for the Solution of NavierSTokes Equations. "An Evaluation of Upwind and Central Difference Approximations by a Study of Recircul ing Flow. and Nakayama. and Lewis.. K. J. C. and Hansen.G.. Covelli. W. G. J.H. Y. "Tensor Viscosity Dukowicz. R." Journal of Comp. Jr. "Coolant Mixing in the LMFBR Outlet Plenum. Academic Press (1964). Phys... T. Clark. "Accuracy of Difference Formulation of NavierStokes Equations. Hirt..B. L. Vol. 7179 (1979).P. Hirsh. 4.. 1 (Nov. Cebeci.D. "Numerical Methods of Reactor Analysis.. Phys. and Design 69. G. Keller." Nuclear Eng.C. and Design 69. thesis (1977).. B.. [14] [15) [16] [17] 118] [19] [20) [21] [22] 123] [24] [25) ." MIT Ph.. P.. Varadi. N." Nuclear Eng." LA6236 (July 1976). J.I. and Eng. and Ramshaw. L. J.. Nielsen.D.D. "VARR2 ." MIT Ph. and Romero. S. Method for Convection in Numerical Fluid Dynamics. "Buoyancy Effects in Mathematical and Numerical Turbulence Modelling. Chang. 1344 (1981)." CRBRPARD0106.K.J. "SOLAICE: A Numerical Solution Algorithm for Transient Compressible Fluid Flows." Comp.L.." Journal of Comp.W..D. "Study of Nonhomogeneous Conditions in Reactor Containments Due to Separation Effects. M. Crowley. of Fluids Supplement 2 (1969). thesis (1980). Vol. 27. 471484 (1967). S.S. Williams.. 1976)." The Phys.F." Computers and Fluids. 127137 (1982)..H. "Simulation of Containment Cooling with Outside Spray After a Core Meltdown. 1. 95107 (1982). M.
101. 18. C.. J. Vol. 2nd ed. Wiley. "Improved Form of the Low Reynolds Number kc Turbulence Model. F..H. "A Comparison of Hybrid and Quadratic Upstream Differencing in High Reynolds Number Elliptic Flows.. (1975). 309312 (March 1975).A. "Computer Studies of Timedependent Turbulent Flows. Molecular Theory of Gases and Liquids.." Comp. 8093 (1968). 15. C. 17. Gupta. J. and Eng." The Phys. "Numerical Calculation of Multiphase Fluid Flow. M. Hoffman. Phys.. and Launder.B.. "Behavior of Buoyant Moist Plumes in Turbulent Atmospheres. Harlow. J..P. B. 2..M." MITEL 81020 (June 1981)..O.A." Journal of Comp. [27] [28] [29) [30) [31] [32) [33) [34] [35] [36) [37) Hirt. Hirt.H.A.H. and Golay. "A Numerical Fluid Dynamics Calculation Method for All Flow Speeds.W. B. for Num. R. 1952 (1975)..The 1978 Freeman Scholar Lecture. and Amsden. A. Phys. "Numerical Calculation of Almost Incompressible Flow.. and Bird.O. Hamza. "On the Use of Central Difference Scheme for NavierStokes Equations. Methods in Eng. Han." Journal of Fluids Eng. Harlow. G. Curtiss. [38) .. 197213 (1971). Phys. 557573 (1980). "Practical Investigation of Convective Difference Approximations of Reduced Dispersion. of Fluids Supplement 2 (1969).H. J.W.. Methods in Applied Mech. J. and Amsden. of Fluids.E. "Heuristic Stability Theory for Finite Difference Equations. R. Vol. Hirschfelder. M. A.. Humphrey.. Phys. C. Hinze." The Phys. "Buoyancy Induced Fluid Motions Characteristic of Applications in Technology . 528 (March 1979). Gebhart.. 8..." Journal of Comp.C. McGrawHill Book Company. F. and Amsden.A.. of Fluids Supplement 2 (1969). R..W. 339355 (1968).E.172 [26] Fromm. Turbulence. T. New York (1954).." The Phys. Vol. F. A..F. 8195 (1981). Harlow. and Marohar. 3. 29." Int." Journal of Comp." Journal of Comp.
J.E. and Rodi. and Golay. Vol.. 3. and Launder.D.H.V. Fluid Mech. Launder." BF63. and Eng. Manno. Vol. translated (October 1980). W. B." J. Vol. "A Modified Form of the kc Model for Predicting Wall Turbulence....E. G. 301314 (1972).. Frankfurt. 269289 (1974). 16. J. Vol. Cl. "ADI Solution of Free Convection in a Variable Viscosity Fluid. and Bremhorst. J.G. Methods in Applied Mech. J. Launder." MIT Sc.W.G.." Comp.E. Heat Mass Transfer. Keenan." Report submitted to 2nd International Topical Meeting on Nuclear Reactor Thermalhydraulics (March 1982).K. [40) [41) [42) [43] [44) [45] [46] [47] Leschziner. "Practical Evaluation of Three Finite Difference Schemes for the Computation of SteadyState Recirculating Flows. Steam Tables." Trans. 11191130 (1973). and Eng. 1981). 293312 (1980). 15.P.. 537566 (1975).. 68. "The Calculation of Low Reynolds Number Phenomena with a Twoequation Model of Turbulence.. Jones. 103." Journal of Comp. "Analytical Modelling of Hydrogen Transport in Reactor Containments. Phys.P." Int.." Int.173 [39) Houston. "The Prediction of Laminarization with a Twoequation Model of Turbulence.P. Reece. J.. C. Heat Mass Transfer. thesis (1983). Manno. V. W. G. "Analytical Modelling of Hydrogen Transport In Reactor Containments. Jones..E.3633. and Keyes. M.A. "Progress in the Development of a Reynolds Stress Turbulence Closure... M. Lam. B. "The Numerical Comparison of Turbulent Flows. Langer. B.. B. V. 16.. et al. of the ASME. Jr.B. and De Braemaecker." Comp. K. 221239 (1974). 148) [49) . and Spalding.P. 23. "Experimental Investigation of Hydrogen Distribution in a Containment of a Light Water Reactor Following a Coolant Loss Accident.H. John Wiley & Sons (1978).. D. 456460 (Sept. and Launder. F. part 3. Battelle Institute e. W. Methods in Applied Mech. M.
P. 4. T." The Phys.B.D. John Wiley' & Sons. Turbulent Buoyant Jets and Plumes.. D. Computational Fluid Dynamics. "Convective Difference Schemes.. Reid.. and'Eng. Roberts." Trans. Phys. 10. Methods in Applied Mech. Numerical Heat Transfer and Fluid Flow. "On Artificial Viscosity. Methods in Applied Mech. and Spalding.. McGrawHill. R. [52) [53) [54) [55) [56) [57) [58) [59] [60) [61) [62] [63) . of the ASME. W.D. Albuquerque. New Mexico (1976). Patankar.. "Convergence and Accuracy of Three Finite Difference Schemes for a Two Dimensional Conduction and Convection Problem. "A Note on the Empirical Constant in the KolmogorovPrandtl EddyViscosity Expression. Patankar. "A Calculational Procedure for Heat.. W.O. 1975). Rodi. 2. Roache. and Torrance.174 [50] [51) Nakamura.. 191206 (1974). 9. "A Critical Evaluation of Upstream Differencing Applied to Problems Involving Fluid Flow. Computational Methods in Engineering and Science. A.J. K. of Fluids Supplement 2 (1969). "Numerical Experiments with a Difference Model for the NavierStokes Equations (Turbulence Model). Raithby. Runchal. Methods in Eng. K. 15. S. 17871806 (1972). Roache. G. "Upstreamweighted Differencing Schemes and Their Application to Elliptic Problems Involving Fluid Flow." Int. 9.. (1977). The Properties of Gases and Liquids. 20 (1966)." Mathematics of Computation.. Rodi.. Pergamon Press (1982). Inc. Vol. 386389 (Sept. and Weiss. 75103 (1976)." Comp..K.C.. N. Mass and Momentum Transfer in Three Dimensional Parabolic Flows. Hemisphere Publishing Corporation (1980). for Num.B. J... G.D. 169184 (1972).. 153164 (1976).J.E..." Journal of Comp. W. New York (1958)." Comp. P.. Raithby. and Eng. Hermosa Publishers.K. G." Int. Heat Mass Transfer. S.V." Computers and Fluids. "SkewUpstream Differencing Schemes for Problems Involving Fluid Flow.V. and Sherwood. J. Raithby. S. 541550 (1972). Schonauer.
" FATE80114 [73] (Jan. 1974). and Strong. R. "A Novel Finite Difference Formulation for Differential Expressions Involving Both First and Second Derivatives. 1981).." SAI74509LJ. Smith. Stubley.. Vol. No.L." of the National Bureau of Science .. 411428 (1980).. Strong. "TEMPEST: A Threedimensional Timedependent Computer Program [72] for Hydrothermal Analysis.H.. "Comparison of Finite Difference Journal of Computations of Natural Convection. 35. J. "The Numerical Treatment of Advection: A Performance Comparison of Current Methods... Spalding. A.. 3. "Numerical Analysis of Fluid Hammer Waves by the Method of Characteristics. and Valentin. 4.D.G.R. 4 (Oct. and Eyler. MatheResearch matical Sciences. 95102 (1950)." Numerical Heat Transfer. Stubley. A. "Diffusional Properties of Multicomponent Gases. . Methods in Eng. 411428 (1980). G. G. A First Course in Tur [65] [66) [67] [68) [69) [70] [71] Torrance.B. MIT Press (1970). Budden.W. L. A. Methods in Applied Mech.J.." Journal of Comp. Y. bulence.A. Tennekes. 72B.D. G.. J. 551559 (1972).... 153168 (1982). 46.. R." Int. D. Wilke." Chemical Engineering Progress.A. D.. Raithby. Raithby. "Development and Validation of a TwoVariable Turbulence Model. Phys.L. Vol.S.Dec.B. 20. 2. K. For Num. No. 220237 (1976). "Proposal for a New Discrete Method Based on an Assessment of Discretization Errors.B. and Lumley.D. and Woolner. Stuhmiller.M." Numerical Heat Transfer.D.175 [64] Shin. 1968). C.B. (Jan. H. M.. K. and Eng.E. G. Trent." Comp.. "Simulation of Convection and Diffusion Processes by Standard Finite Difference Schemes and by Influence Schemes. and Hutton. 3. J.
2) Equation A.a2T aT 2 + a2 2 2c ax ax 2 y ay (A.2 is simplified further by substitution of 4 for T as follows. 5. The boundary conditions are specified in Fig.1.1) In Eq.1 convection occurs in the x direction and diffusion occurs in both x and y directions. A. Therefore.176 APPENDIX A ANALYTICAL SOLUTION FOR THE PROBLEM IN FIGURE 5. as follows.3) ax . A. 2 x 2 a2 ay = c 2 (A..3(B) A steady state two dimensional energy conservation equation without any source is given in the follbwing Eq. Eq.3(B) and the constants can be grouped into one constant. pc u aT = k(a2T + 2aT 2 x2 p x ax ay (A. c. a= k PCp C =2 u 2A.1 is2 reduced to the following form.
c 2 X cx 2 = (A.7) d Y c . I d2X + 1 d 2Y 1 = c2 C (A.a =  (A.3 is reduced to the following form.4) The boundary conditions should also be expressed in terms of 4 as given in Fig.5) Then Eq. ables. A. A. which are functions of x and y only respec (x. A.8) dy where a > c > 0 .1.177 where 4 = eCXT (A.6) dx dy The two terms on the left hand side of Eq. Equation A. tively. The variable 1 is separated into two vari X and Y.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.y) = X(x)Y(y) (A.3 can be solved by the method of separation of variables.
4 = ec T .0 0 = 0 x Fig.1. 5. A. Boundary conditions in terms of 4 for the problem in Fig.= YO T 0 ii ii i h YO + co = 0 4 = 0.3(B) where 0 is given by.178 .
Eq.9 and Eq.7 and Eq. A. A. the constants in Eq. X = Cleax + C2 eax (A.179 The solutions to Eq. A.1 may be given in terms of the variables X and Y as follows. The boundary con dition at x=0 will be considered later. A.11) dY dy=Y 0 = 0 (A.16) . A. 13) In order to satisfy the boundary conditions given in Eq.12 and Eq.12) (d dx _cX) = 0 x=xO (A.10) The boundary conditions in Fig.10 should satisfy the following relations. dY dy y=O Y=O 0 (A. A.14) (A.9) Y = C3sin8y + C4cosSy (A. 13. A. C3 = 0 (A.11.15) ny 0 = n C2 _ n+c 2 e nX 1 n (a n 9 C) (A.8 can be easily obtained as follows.
Th 0 2.1 are determined as follows. . the final solution can be expressed as follows.y) where nr (A.2. 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.180 where 2 =c 2 + n (O) n 2 yO =c 2 + 8 2 . nw sin2 nO0 (n=1. .18) n n Since the boundary condition at x=0 should also be satisfied.3 . 17) T(x.) (A. the constants A.0 2 A n Th h nr 1 n 1+( a c )e n 1 a +c 2a x . Therefore.y) = e cx (x.19) .
the time step size in the ADI should be at least two or three times the Courant limit in order to have a gain in the overall computational efforts required for a problem. B. there are six sweeps in a three dimensional problem and four sweeps in a two dimensional problem.1 and B. The time step size limitation in the ADI scheme comes from the semiimplicit treatment of the nonlinear convection term.4.3 are the z and x direction sweeps of the x direction momentum . Equations B. the computational effort for one time step is greater than that of the explicit scheme. When the tilde phase of the SMAC scheme and other scalar transport equations are solved by the ADI scheme. The ADI scheme is applied to a two dimeraional momentum equation in the following Eq. B. Since there are two steps involved in solving the momentum equation in each direction.iillaY~i~~l_ ^YII^^~II ( . 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.1Eq.i. They are formulated for a natural convection problem where the gravity acts in the zdirection. The sequence of the sweeps is arbitrary and depends on the problem under consideration. The most important sweep should be given the first priority and most updated information. Therefore.
2ui+. ar .1) * * n * *  ij+ At Wi+1 j+ ij+ + i+Ij+wij+ uij+Wilj+ 1 P n~ .182 equation.u n+l ui+2 2 n+l Ar2 +2u j n+l ui j .P z  n * RZij+ ij+ ij+ (B.2 and B. Equations B.u Lr * n+1 . u n * u j i+ Ui+D Dt j + * n i+ j+ Ui+ j .Wi+ Az j Ui+ j 1 ui+ j+l * .3) . sweep of the z direction momentum equation is treated as the last one in order to be given the most updated information.2) n+l i+j * .Rxi+ jui+ jui+ j RX * n+l (B.4 are the x and z direction The z direction sweeps of the zdirection momentum equation.ui+ j At  n+l * u i+lj i+ .2ii+ Ar 2 + Wilj+ . 2 Az + Ui+ 1 P  n  i+j i+ jui+4j (B.
+ n+1 n+l Az 2wi + w _P z P0 zP g z RZ n+l ij + wij+ 3wij+ (B. B. B.4) Equations B.183 n+l i+ * wij+ * i n+l +lWij+ * n+l zW. and B. w At n+l Wi. .3.1.4 have given almost identical results with the explicit scheme for a time step size below the Courant limit.2.
are the computer programs for calculating the numerical values of the analytical solution in the problem geometries. Then four crossflow diffusion constants at every interface are summed up to give a total crossflow diffusion constant . All the inflow velocities The component velocities are used to calculate the crossflow diffusion constants for eight subspaces in a three dimensional coordinate space. a flow split occurs as shown in Fig. There are When there are inflow and outflow. The file.3(A).184 APPENDIX C COMPUTER PROGRAMS The files. two different cases treated separately. 4.8. or all inflows in x.3(B) and Fig.3. When there are two outflows in any of the x.8.. 'prog2 fortran' and 'prog3 fortran'. are assumed to be zero. 5. 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. is the program for testing Fig. y and z direction mesh interfaces. 'progl fortran'. y and z directions. 5.2. 5. 'difprg fortran'.4. Fig. the calculational logic of the mesh interface implementation of the corrective scheme in section 4. the crossflow diffusion constants can be calculated in a simple way as in the first part of the program.
. A portion of the VARR program is introduced to show how the ADI scheme is implemented in the energy and momentum equations. The subroutine 'tdm' solves tridiagonal matrix problems by forward and backward sweeps. The total crossflow diffusion constant can be calculated in the same way for every mesh interface. has four neighboring subspaces. The ADI scheme for the momentum equation is given in Appendix B. It is used to solve a one dimensional implicit finite difference equation in the ADI scheme. the positive x direction.185 because one axis direction.
26 II IF  IP  I GO TO 55 ALN = DSORT(CSO*II*II*YSO) (II.04ALP) CSO .EO.14 55 ATOT * THIODEXP(ALN*XO)/2.I=l.YO) CCOrrF 2.NY. C*C XO " NXOX YO * PROOOO0 PROOOO20 PROOO30 PROO0040 PRO000OOOO P14000070 PRO000OGO P11000090 POOOOT100 PROOOO10 PIIOO0120 PRO00130 PROOO140 PROOO150 PRO00160 PROO000170 PRO00100 P1.O) ALNX .COF*DSIN(II*HPI)/(1.3F7.4ALNX) ATOT .0 00 20 IPsi.N FILE: PROG FORrRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT REAL*0 (AH.3) SlOP END PRO00210 PRO00220 PROO0230 P1ZOOO240 P1I000250 PROOO260 PROOO270 PROOO200 PROOO290 P11000300 PROO0310 PRO00320 PRO00330 PROOO00340 PRO00350 PROOO360 PROOO370 PROOO300 PROOO390 PROOO400 .IALNIC)*DEXP(2.O*ALNXO)/(ALNC) COF * CCOFF/lI AAN .5)/NY ADD * ATOTTDCOS(YY) SUM T(d) v SUM + * ADO * DEXP(C*XO) 20 CONTINUE SUM 10 CONTINUE WRITE(G.NY  0.OZ) T(tO) DIMENSION REAO(10.ALP II FOIRMAT(F7.2.2) PI a 3.5F11.NX.0*TII/PI PI/2.NY) 100 FOnMAT(IOX.100) (T(I).1 ) DX.215.0 44 YY I IIPI*(UJ0.0 IHPI DO SUM 10 JJI.U.XO)/(ALNC) GO TO .OALNOEXP(ALN.1415926536 C = U/(2.000190 PI00O200 NY*OX YSO v PI*PI/(YO.Tl. AAN*2.
0 SUM3X 0.NY2 0. NY.O.5*YO .X = 1. IC1l REALi0 (AII.0  RTO DO =  OSORT(XXX) 10 JJ=1.0 .U/(2.DCOS(BEN*YI) YI Y22 • DCOS(BENY2) OCOS(BEN*Y3) Y3C Y3S .2. ALP READ(10. 11 FORMAT(F7.FO.1.30 P1OOO 140 POOO00150 900440 PROO0170 PRO00180 PROOO 190 PROO0200 PROOO0210 PROO0220 PROOO230 PRO00O240 PRO00250 PR11000260 PRO00270 P1O00200 PR000290 PROOO300 PROO0310 PR000320 PROOO330 PROO0340 PRO00350 PROO03GO PROO0370 PROOO300 PR000390 P11000400 PROOO4 O PROOO420 PROO0430 PROO440 POO450 PI0OO0460 PR000O470 PROO0480 PR000490 PROOOSOO PR00510 DX'NX YO " DXNY VSO " rPI*P/(YOYO) CCOTF = 2.0 = YO/2.0 SUM2 = 0.0 = 0.3. TC(IO) TH14.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN .+ALNX) BEN 11*PI/YO YOO .O) GO TO 55 ALNX .215.0 SUMaX SUM4 XN vI Y2 Y3 Y4 • 0.0Z) DIMENSION TII(10).0 NY2 XXX NY/2.5) XALO = DEXP(ALN*XO*O.3F7 .0 SjU3 " = 0.5) EXAL = OEXP(ALNOX) XAL = DEXP(ALN'X) EXALN D EXP(ALN*XN) XALN = DEXP(ALNXN) IF(II.0 X + XO/2. NX.p A FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 MPI.COF*DSIN(II*HIPI)/(I.5+YO 00 20 IPs1.OALP) CSO = C'C XO * PROOOO O PRO00OO20 P R000030 PROOOO40 PROO00SO POOG0050 PROOOO000 PROOOOOO PROO0090 PIR000 100 P11000110 PRO00 120 PROOO 1.DCOS(BEN*YC) . 1415926536 C .5)*DX = YC = (JJ.(ALN#C)*DEXP(2.21)*DX = X 4 YO/2.11) DX. TL(10).OCOS(OEN*Y4) Y4S DOSIN(3EN'Y4) ADO * AAN*(EXALO + ALNX*XALO)*YOO PROOO520 PR000530 P1OOO00540 PR000550 . U. GH(10). GL(10).2) PI .0 SUMO SUMI r 0.0 X = (JJO.0 .OTHIH/PI 11PI = PI/2.DSIN(BEN+Y3) Y4C .XN = XN .21 II = IP  I ALN = DSORT(CSO + IIII*YSQ) EXALO = DEXP(ALN*XOO.
AAN*(EXALALNX*XAL)*Y22 ADO3X v AAN*((C+ALN).5) SUMI*DEXP(C*X) SUM2*DEXP(C'X) (SUM3X+SUM3Y)*DEXP(C*XN)/RTO (SUM4X+SUM4Y).0 ADOvY * 0.NY2) = I.00) (GL(I). 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.(" % FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 ADOI w AAN*(EXAL+ALNX*XAL)*YII AD02 .0 ADO3Y * 0.II.SFII.DEXP(COXN)/RTO PROO000560 PROO0570 PROOOS80 PROOO590 PROOO600 PROO0610 PROOOG0620 PR000G30 PROOO640 PRO0650 PRIOOOG60 PROOG0670 PROOO680 PROOOG90 PROOO700 PROOO7 O PROO0720 PROOO730 PROO00740 PROOO750 PROOO7GO 1PI000770 PROOO700 P000790 PROO000000 PROOOIO1 PIOO000120 PROOO030 PROO0040 PROO000050 PROOO060 PROOO870 PROO000 PROOO090 PRO00900 PROO0910 PROOO920 PRO00930 PR000940 I0760 00 oo I .II.5 A002 * TMIIIXAL0.NY2) WRITE(G6.II.tOO) (TL(l). 1OO) (TH(I).5 ADO3X 0.5 A)DO * TIHl'XALO*O.EXALN + ALNX*(CALN)*XALN)*Y3C EN*AAN#(EXALN .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(6.0 AOOX a 0.I WRITE(6.NY2) WRITE(6.tOO) (TC(I).NY2) WRITE(.3) STOP ENO SUMO*DEXP(C*XO0O.NY2) = t.100) (GH(I).I 1OO FORMAT(IOX.
2. G1(10).(dJ0.215. TC(1tO) PROO0020 PROO0030 PROO0040 PnOOO050 PROOG000 READ( 10.O' XO/RT3)/NY2 XI=(dJ0. rORMAT(F7.NY2 SUMO = 0. ( 10).0*RT3))/NY2 Y3=(JJO.5)*(XOYO/(2.0*RT3#NY2) Y2=YO/2.0 SUMI = 0.0) DO 10 00=1.22 II IP I PR00360 PRO00370 PROOO3UO P(1000390 P000400 ALN = DSORT(CSO + II*laIYSO) GO 10 55 Ir(II.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 .OtTIIIM/PI HIPI = PI/2.O*NY2) .0+(dJO.0.0 SUM3X = 0.0(JJ0. FILE: PROG3 FORIRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT DIMENSION C 11 REALtO (AH.0 XC=Y0/(2. ALP PROOOO10 GL( O).5)*(XO/RT3YO/6.0 PR00O170 PR000100 PROO0190 PR000200 PlO00210 SUM2 = 0.1415926536 C = U/(2.*0RT3#NY2) Y4=YO/3.RT3)+(dJO.4) Y4S = DSIN(6ENlY4) ADDO=AAN*(DEXP(ALN*XC)4ALNX*DEXP(ALN*XC))*YOO ADDIAAN9(DEXP(ALN*XI)+AI. TIIl. U.S)*(XO/RT3YO/6.2) PI = 3.0) PROO000120 PROO000130 PROO0140 PROOO150 PR000O160 RT3 = C SORT(3.C.0 SUM4Y = 0.0 NY2 nRo P000O070 13100000 PROOOOOO PROOO0000 PRO00110 = NY/2.o) ALNX = (ALNwC)tDEXP(2.DCOS(iEN*tY2) YV3C = DCOS(BEN'Y3) Y3S = DSIN(BENtY3) YAC = OCOS(3EN*Y.02) 111(10).+ALNX) OEN = II'PI/YO YOO . NY.XO/RT3.5)*(XOYO/1r3)/NY2 YC=(JO.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN = COF*OSIN(II'HPI)/(I.0)/NY2 X4XOYO/(2.5)YO/(2 .0)/NY2 X2=(JJ0.0*RTJ))/NY2 YIYO/2. I 1) DX.0.NX*DEXP(ALNXI))*YI I ADD2AAN*(DEXP(AI.0*ALP) CSOQ C*C XO = DX'NX YO = DX*NY YSO = PI*PI/(YO*YO) CCOFF = 2.0 .EO.0 SUM3Y = 0.5) * (YO/3.0*NY2) X3YO/(2.DCOS(UENOYC) YIl DCOS(OEN'YI) Y22 .O*RT3)+(JJO.0 SUM4X = 0.5)*YO/(2.C P11000220 PROO0230 PROO0240 P1000250 PROOO2GO PROO0270 P11000200 PROO290 PR000300 PROO03 10 P 11000310 PROOO320 00 PRO00330 PR00OO340 PROO0350 00 20 IP=1.5)*YO/(2.3F7.0*RT3)+(J0. NX.5)YO/(2. TL(IO).SO5T(2.S)#(XOYO/(2.
I= . PROOO6tIO .5 ADDH11H*DEXP(ALN*XI)*0.11.I=.NY2) WRITE(6.NY2) WRITE(G.5+SUM3Y*0.NY2) tOo rORMAT(IOX.0 ADO3Y = 0.100) (TH(I).5*RT3)*DEXP(COX3) GHi(JJ)(SUM4X#O.5)*DEXP(C*X4) 10 CONTINUE WRITE(6.I=I..OEN*AAN(DEXP(ALN*X4)4ALNX*)DEXP(ALN*X4))*Y4S GO TO 77 55 ADDOTIIII*DEXP( ALN*XC)*O. = 0.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 ADDOX a 0.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.X4) ALNX*(CALN)*OEXP(ALN*X4))*Y4C ADO4Y.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. 100) (GI(I).SF11.5 ADO3X PROOOSGO PROO00570 PR1000500 PROO000590 PROOOGOO P1100000.5 ADD2TtII#DEXP(ALN*X2)*O.1OO) (TL(I).100) (GL(I).100) (TC(I).NY2) WRITE(6.0 ADO4Y = 0.I1.3) STOP ENO .NY2) WRITE(6.5*RT3+SUM4YtO.0 77 SUM1 .
O.O.EO.0 ) C DIFR .VP.) ISA .ISC.5*( VM V2 0.0) IF(UP.L1.5.0.VM.O) IF(WP.5*( WP C IF( ISA.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 .5*( UP VI 0.UU/DFLR PPZ * VV/OELZ PPT .0) IF(VP.4 ~ FILE: O FI'IlG FOITIAN A VM/SP COINVERSATIONAL MONITOR SYSTEM PACE 001 DIMENSION VELX(0).2.VELZ(0).EO.ISt + IS2 ISO IS3 + IS4 ISC ISS5 + ISG C UI = 0.0.GT.5+( UM U2 0.O 15S3 .UP.WM.0) IF(WM.VELT(O).VI + V2 WW = WI 4+ W2 PPX .O.WW4DELTII#( PPX + PPZ )/PSUM C RiFGII .0 IF(UM.DIFTA(0) REA( 10O.0 n I FOO 130 OMTI m 0.UU*DELR*( PPZ + PPT )/PSUM DIFZ * VV*DELZ*( PPX + PPT )/PSUM DIFT .2) UU * UI + U2 VV .DIFRA(0)1.O.DIFZ'IS3 * DIFZ*IS4 RFGr GO TO 00 AOS(UM) + ABS(UP) DI F00170 DIFOO IGO DIFOO00190 DIFOOlOO ISI IS2 IS3 IS4 155 ISG .0) IF(VM.PPX $ PPZ + PPT GO TO 500 IF( PSUM.OR.0 RFGTI . bt ts .t = 1 = i v 1 * .LT.DEL'TII.DIrn*ISI .5*( VP WI * 0.( WM W2 0.0.0 RFOT 2 0.0 RFGCR 0.OR.0 OMT .I 01o00200 DIF00210 DIF00220 DIFOO230 DIFOO240 DIF00250 DIF00260 DIFO0270 DIF00200 OlFOO290 DIFOO300 DI F003 10 I'I .O.DELZ.GT.5) DELR.EQ.GT.IS .2.O.WP 1SI * 0 152 .O 151 4 0 ISS1 O 156 O DIFOOOIO DIFOOO20 DIFOOO30 OIF0040 011:O0050 DIFOOOGO DIF00070 DIFOOOTO DIF00110 RFGRI = 0.WW/DELTII PSUM .IFZA(0).UM.DIFII'IS2 rGR RfGTI .LT.EO.
0 oiETA(I) .0.0PFX)*PFZ DO 30 11.( i.0PFz) W2 &1EX'&( 1.O.0PFT) V t.0PFT) U2*PFZ*( 1.0 oirZA(l) .VEU.0.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$( t.0PFT) U I PFZ (I 1.*( PPX + PPZ )/PSUM GO TO 30 600 OIFRA(I) .o 'FX'( 1.V(LT(I).EQ.OPFX)O( I.0 Ir (W1#W2).PrT )/PSUM DIFTA(I) .09'FX)#(t1.0.0PFX)'PFZ Wt'(I.PFX*PFT V2* I'F X'PF T V2.0PET) Vl*( 1.0..FILE: DIFPRC'.X(I)'DELR*( PPZ + PPT )0 'PSLJM DIFZA(1) m VELZ(I)*DELz#( rrX +.0 ) GO TO 600 DIFRA(I) .0 PPX .( .) U2 4PF Z PET 1J2'( 1.( 1.NC.0.VELZ(I)/DELZ PPT .0PET) Vi*PFX*( 1.0 Pr T Ir (UI+U2).DELTH.VELX( I)/I)ELR PPZ .0OPF Z) W2 *P FX' IF Z WI'.0PFZ)*PFT U20( 1. oi rooG to Dl F00G20 D1F00630 01F00640 Ut 'PFZ .0PFZ)'$( .VELT(I)/OELTH PSUM a PPX + PVZ + PPT IE( PSUM.0 0.0 (Vl*V2).NIE.NE. 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.IPFZ)s( 1.oPF7Z)'PFT UIs .0PFX)#(t.0PFZ) W20( 1.0PFX)#PFT V2( 1.PFT) WI#PI'X#PFZ WI'*ix( .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 .0PFX)o( t.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. 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 *0.0PFX)*PrET V2*( 1.0PFT.0PET) Vt .PFT u i.0.
6F5.OMT I.20X.RFGT.2X. 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 .RrGT I.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 .RFGR.I.2) 7 FORMAT( ItII .2) G FOIMAT( III .7) 5 FORMAT(9F5.OMT GO TO 200 500 WRITE(G.
X(N).O(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.0t4O01 X( I ) I~) 13( ) 0(i1) NM I *N I 00) 2 1I.D.o(N).0.SUarIouT INE TDM(A .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) .NMI .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 .C(N).P(400).C .
xx'.CR.O.G.KBP2 " (3.:X GO TO 40000 "0OO D.VGRD2/(.)K2NC " (Ki)'NWPC .UG.K) • AVG.F2(J. 1) GO TO 400 IF(K.KZ2))) .K t KXX a jK .KS 2.S(0.L.JSP2 DO O40000 DO 4000 K.]K .C .C "hT.E0.0.E230 .LT.UG.0 OOC COI~'T2]'UE C VAR P182 vLR a210 VAR E220 VAR.G  V RIS150 VAR.DFFUSI'O =.FX(3..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.P2) GO TO 40002 (U(iK)*U(KX))) LUG = LS(O.ItD.I..FFX ( .S( 4 . I O50 VI8090 K1 K12 2 KEP CONSTA'NT CLLCULATION C CRO5SFLD'. DNF. I vtR I5S0 rooOT 00G2 UC 'S(U(]K) ) VLRi 1210 vAR1E230 VAR E20 U(3KXX) ) LUGc . C C C C U AND v LDE VELOCITY EDU.S ..K2NC JKZ2 .0 D'rF2( I .K) * 0.) G0 TOrOOiC D'.K) z 0.TIONS SECT ION AD !ST ]F(ICLL].5(V( K)*:(. 1CONV..P.1) GO T0 49998 SOLUTiON FOR SIE.3.I.L!S( (]KZZ) ) .PM.EO.LE0. D2).LT. iB10 .2D LUGRDX/(AUGOX +~DG'RD)' .E1O 0005 FP" * IF(...K) a LUGDX.NPC o1 ]F(K.E0. DA'' G.EO.
( '(K ULH C CON'.1.. VLRIE6t0 V4RI15650 VAR 15690 ULWM'(uLLIC) ( IR'(L'C4UV) VARIE *7 DO. KS I KLS a3)K I KP a IK  NP 2 I N1PC VAR 185501 YLR 15560 VAR 1C570 VARias590 VAR 86201 VARi1Es30.P( TJ"'.]BR ( '.0.E.KSPI I KL I lK z lK 1A 0.R2CLI.S(UftH)(UCUQ)  VL21S660.5 UODUMKS) 16D(1PK) * .2)GO a TO 29990VALR iS.E003 DO E0100O ) K GO TO50205 222.5 )r0.ET/TADD3 VAR IS420 VLO IS440 3F~lDw.I) K2WC' + II)K NC *K2NC "00Z0100 K=2.C RD20 3KS0 S3G'~ to VAR 8340: VARIS5360' RDZDRD2 VAR i !:0.'O('.I ) I KIPC # (KI)'tN'PCL VAR16D VAR 11510 VARi18520 !PK a 31. a uJD(2K) )0. 054 (K.PKKNUO(IPK) )..I DIFFCD(3KL2) D!~FFCD(IKL3) *0(!. .0.1 SRKSrI a * N BUOJY a LP0E7.5 PC) '). 3F(3~.. UD( 3K) UL UR UT Uo PK) UD( 3KP) U0( 3KMS) DCR DCL D!FFCD(IKL) D) F CO( KL.
t) 2 td9N'Zt oo0to oa 3flN tLNOD s0?_O3 M 6 dVA + NAcns + ZntO * (Nt)d(dXt)d ).Cw'))'A.(onin) M0'a(On.=OStd7A 0506LdV  ~ ~ ~ ~ Hm )CC7r~! Sall*~ ZN07J ±.% At t x NdE ONCA + lA b + A t + OZ~N(Lt) OdAN 00gla aVA 06!.vt Z1t0 7 Z(J~Si0(o(nsn)(a)a (1n.uQ + + t (Nt)0ftl x = XIt)f ( oo..onH. (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'.) .o2m.OSZ6 M7A OdI'N Al dN O7I~t 17A OdM ( A) ONZ Nt + 1IOZN4(Vt) * + 3NZ)(11) + t bi > >4 E A c TA .8~ VA 0899tMVA (oaQN 3NC647.'N(tX) + 06Md~7A 1 Et a 6Nr NZ At 2 SW SWA r r ). Qti 0*0 o3 a itflt NACAS dsm~rsN td3t*Z9t Oozos 00 00ZO9 00 0LL8Z!VA 0SL8dVA (Zito ZNOO YNOO0 YO + ) 00L3'ttVA 06L9tMVA 0 LStMVA (At)d(Adt)d  11 )*.(?nOn) i13( fl2f) unc.( (nf1l)(Himlss7 (In3n)(uLM)S9V ) ZNQ..+fl)IdM Y(Qd.o.)10 6666a 01 0 MfN i NGO 07 6 "VA ZNOO NCO ZOM( + (Xt)0M £ ) O~t6tVA (~ ~ .(.(iim1 (..7A 0060tMIA 0L6Sgv7A 0968LVA I 3a CA~ SimI) 0L96 t'7A O9S63tM7A j0 0 VA d061Nt) AoflsaNx0As + Od.
ABS(ULT) ) .5RD02( WRe .5RDX( URT 1 + SIGMARDXD.'5 WRS (W(1KMS) * W(IPKNWPC))O.IKBR) CALL TDM(A 1.IBR DO 81100 K2..5*RDX( ULT .NWPC DCT 3 DIFFCO(IKL+1) DC.KSPI ]K = 1 + (11)K2NC + (KI)NWPC 1X a (12).5( U(IK) + U(]KP) ) ULT C a 0.5.X.]K .VRE + ABS(WRB) 1 * SIGMA*RDZD.KBR) C DO 82100 Ka2.$ ]K .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 .ULT * ABS(ULT) 81(J) .16PI 1K a 1 * (]i)K2NC + (K1)NWPC IX x (K2)ISR + I 1 E2100 W(]K) * x(Ix) 00 823000 K2.C1 .SIGIArRDZD3DCT 01(J) a U(IK)ROT 81000 CONTINUE C .1 ]K £ + (1])K2NC + (Ki)NWPC ) VAR 19260 VAR 19270 VAR19 280 vLRlg2O VAR 19290 VAR19300.( DCT + DCS ) C1(J) a 0.DIFFCO(IKL) DCL a D0FFCD(IKL+2) URT x 0.KBR DO 62100 1=2.KER D00 82000 12.( DCL * DCR ) C1(J) z 0.0.C1.K2NC .ABS(VRT) ) .W(1K)RDT 82000 CONTINUE C CALL TDM(A .SIGMARDXD*DCL ACBS(URT) .AeS(WRB)) .'RDXDDCR D01(J) .SIGt.( URT EAS(URT) ) .1SP J a (K2)]IeR * 1 .5RDZ*( VRT * ASS(wRT) .]K + K2NC IKP a 1MKS 1K + NWPC ]K K K2NC NWPC 1KIAS = C DCR .X. = DIFFCD(IKL+3) WRT 3 (W(3K) + W(iPK))O.DI.ROT * 0.5*RD2( WRT .5( U()]KS) * U(IMKS+NWPC) ) Ai(J) = 0.IMKS IKt. C 00 l8100 I=2. VAR19310 VAR19320 VAR19330 VLR193AO VAR19350 VAR19360 VAR1S370 VARie380 IKL a C 1 + (1)'K2NCL + (K1)NWPCL IPK .5 AI(J) a 0.8e1.KBR + K 81100 U(IK) a X(IX) C DO 82000 K=2.RDX.SIGMhARDZDODCB E1(J) = RDT * 0.
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.~tSC(Hifl)ss b3(3 ).Htin 0~Q0~*/'.' ). )'ZQdS50 )Zc!S*  a ' a  (HS.3nwNtJ. 3)3QO.v4)tS ~m * io1.mNtAz + + ONA(lt.40 C0~7A O CO0d7A NAoC1s (A(I)a .fl.~) Al NP 1=I000ca 00 ..)ssv .0xCd.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( .(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.v%4!)iS +.'%)SS'7 + HSA . z (r 01)~S~7~). ).LG!1CE)PA HIM.(HL')S.Q ( (S'feXtiM 094=J7A OSZO0t7A (>I).ZC(3aE1~S (d~t)d  00oo' ((HIiA)SS7 I )'ZQZ .L 0.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 ).
SIE(IPK) S*EL .5(TS(]K)+TS(lMKS)) VAR203G0 VAR20370 VAR20380 VAR20390 VAR20400 VAR20410 .HDX RRC a I.SIE(IKP) S]EB * SIE(]KW.DASS(UC) AUL . IBR(K2) + I ' (1i)K2NC IK = i + (K1)NWPC IKL a 1 + (11)K2NCL + (K1)NWPCL RC a FLOAT(I1)bDX .0/RC OR = CYL0.2.U(JMKS) vW = W(IKMS) AUC . VAR20420.5(TS(IK)+TS(IPK)) .12 J .NWPC CFR = CF(1PK) CFL .K2NC ]K IK 7]KS IKMS = IK .K2 DO 60000 13a1.CF(UK) IF (CFC.1 (12)(KER1) 84100 W(JK) a X(lX) C 89999 CONTINUE IF(ICONV.1SP1 00 84100 K2.NE.SIE(IMKS) SIET .IKBR) DO 84100 l=2.EO.EO.25*RRC CYRX = CYLO0.CALL TDM(A .1CALI.KSR IK IX a I + (I1)*K2NC + (Ki)*NWPC + K .Ci.NSEO) 33333 CONTINUE 15E0 a 0 DO 60000 K=Ki.S]EO(]K) TSRA TSLA .O.B.DIFFCO(1KL*i) DCL a D!FFCO(1KL+2) DCB * D]FFCO(JKL*3) ]PK )KP a a + K2NC * NWPC = 1K .0P.DLeS(UL) ALC = DAES(WC) 1wS * DASS(WE) = SIE(1K) S!EC SIER . 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 .D1.DIFFCC(IKL) DCT .0.I) GO TO 55000 DCR .CF(]MKS) UC c U(IK) WC = W(!K) UL .S) SIECO .5RRC*RDX CFC .1) GO TO 2040 ISEO = ISEO + 1 IF(ISEO.LE.X.
(nN*31. 06C Zd7IA ZltrEzf 00 OOOL CO 0SCtLb7A NOI0t1D3JCZ NE d33MS CNZ 0L~tZZ7A LCECE OZ :tZ7A 1 05 3 3lNTI.L. t :) a .At~VA OSZI~7A clt'L fa ZAVtAZ'4 0=0i9 00 0=O9 00 I .1! Al z(IJ)L3 (NOfl'Ol V(ISJX017) Ot)t0dVA 0001ZlA N7 dlxiol ((L1)x~JNG4(IV~~sNQ)ts0 . C0 r~t (r)LC 00~ttIdVA 060k ~wVA 0OLt dVA 00009 01 13t5..I vo 0009 0:9>4 I )c t a.(r'.NOD (L03e.C*~ t )4* .x)j! 2(AI1)3rS oo0t 9 06CI ~1VA' 09ZI dVA 0LI.± 4:7) 21J3 1u L7S1L'q'l OS 60ZMVA (Slx () I s I)SO I ((diE)l+ m1)l)'c = vzs I VIL .OSrlth7A OV7 t!VA t.~t J '~1f ((t0AX'1ZJANO ) C dt0* ) * N ) S 4~ I .Q ~ 000 fl co(3 S6 .(N t0 *090~bVA 0O0I bVA 0O'0t MVA OSOI aVA 0OP0tZSVA (Q QtlY."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.+c a 21J10 to.OlI)03(r v+n . t7) . 10 + ( 0S669 01 00 (V03VIJfl S0. N *:dM o zi Z)q4s)4 coo 0.
C2.LE.0 E2(J) .5(DNFFZ(I.IF C DCR OCT DCL DC8 C ( CFC..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 .C2(J)SIE(IKP) C2(J) = 0.K2NC IKWS VAR21460 VAR21470 VAR21480 VA021490 VAR21500 VAR215 O VAR21520 VAR2rS30 VLR21540 VAR21550 VAR21560 = ]K .0 70000 CONTINUE C CALL TDM.0 GO TO 70000 65500 C2(j) E82(J) .NE.5(DNFF2(3.K.0 69950 IF(CFT.1) GO TO e9950 Dl(J) = 01() .5 0.ND.(A2.RDZ C2(J) = 0.0.K)*DNFF2( .k)+DNFFZ(1.I) )XMX GAMTTSTA VAR21620 VAR21630 VAR21640 VAR21650 VAR216GO0 VAR21670 VAR21680 VAR21690 VAR2I700 VAR217 tO VAR21720 VA.0 DI(J) D(J) + DIFFTDCT*RDZ GO 70 70000 65000 01( ) = SIE(]K) L2(j) a 0. 1 * 0.OlFFTDCTRDZD IF(CFB EO.DIFFTDCTRDZD C2(J) a 0.5(WCWC).TTSPLL .IKBR) C DO 71000 ]=m1.01.CO(IK)0.NU) GAMT * RPRAN .X GtA.NNOPT.1.0.X.5(TS(]K)+TS(3KP)) TSeL .EO.12 DO 71000 KKI.5"(ws*Aw )'RD2 .DIFFEDCRD2D RDT * (DIFFTDCT+D)FFS*DC6)*A02D .DIFFCO(IKL) = D]IFFCO(IKL+1) = DIFFCO(]KL+2) .5(rS(lK)+TS(lKMS)) C GAT DIFFT DIFFB C C * RDTSIE(]K) .K+1))X.E0.1) GO TO 65500 D1(J) * 01(J) ..EO.CF(IKt S) WS W(IKMS) WC a W(IK) VAR21570 VAR21580 VAR21590 VAR21600 VAR21610 AWB a DLeS(we) AWC DOASS(WC) TSTA a 0.0 C2(J) = 0.A2(J)SIE(3KMS) A2(J) • 0.1) GO TO 65000 ..M * * IF(TS(]K).I DI(J) a TGA. .(WC*AwCWS*iWB).11.B2.NWPC C CFT = CF(IKP) CFB .i) GO TO 70000 IF(CFT.5.K2 Ix = KS.RDZ .0.DIFFCO(IKL+3) 1PK a 1K + K2NC IKP a !K + NWPC IMKS a 1K .(12) + K .
K2 1K 1 + (]i)'K2NC + (KI)*NWPC 33338 SIE(K) = 0.NE.11. 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.5'( SIE(IK) + STOR(I.O) 00 33335 I11i.K) = SIE(IK) 333335 SIE(IK) .IK r 1 + (II)*K2NC + (K1).E0.SIEO(IK) KIRK a i GO TO 33334 33337 CONTINUE DO 33338 1=1112 DO 33338 K=KI.12 DO 33335 K=KI.K) ) KIRK a 0 33329 CONTINUE C C NOTE. 12 KK=KK4K2NC KKL VAR22010 VAR22020 VAR22030 VAR22040 VAR2 2050 VAR22060 VAR?22070 VAR22080 VAR220s0O VAR22100 VAR22 110 VAR22120 VAR22130 VAR?22140 VAR22150 vLR22160 VAR22170 VAR22180 VAR22190 VAR22200 VAR222A0 VAR22220 VAR22230 VtR22240 VAR22250 VAR22260 VAR22270 VAR22280 VAR22290 VAR22300 VAR22310 VAR22320 VAR22330 VAR22340 2109 VAR22350 VAR22360 V R22370 VAR22380 VAR22390 VAR22A00 VAR22410 = KKL + K2NCL VAR22420 VAR22430 VLR2240 VAR22450 LWPCL = LWPC=I 1 VAR22460 SIE( 1)=S [(KMS) U(1)=U(!KMS) w( ) (IKVS) 0(l)TO(IKMS) TS(i)=TS(IKMS) SIE(]K S)=SIE(KK4+ ) U(KtMS)U(KK+1) W(IKt.K2 IK v 1 + (1I1)K2NC + (K1)*NWPC STOR(I. TRANSFERS VELOCITIES TO STORAGE ARRAY( 2040 KI1 K2xKBP2 LWPC=I .1) GO TO 33333 GO TO 33339 IF(ISEQ.NwPC 71000 SIE(]K) = X(IX) C IF(KIRK.S)=W(KK+) TO(KMS)=TO(KK+1) VAR22470 VAR22480 VAR22490 VLR22500 VAR22510 VAR 22520 VAR22530 VAR225ZO VAR22550 .K2 LWPC=LWPC+NWPC AT TIMEN ) .NWPC DO 2.09 KzK1.
'inputd HUH' and 'inputd DM'. are for calculating the numerical solution by donor cell scheme in the problem geometry. The file. The output The files.6 and Fig. 5. . results are in Fig. The files. 'input mom'. is for calculating the natural convection flow field in a Cartesian closed compartment. 5. 5. is the input to test the stability of the ADI scheme for momentum equation.7. 5. The output results are in Fig.8. Fig.3(B). 'inputn dat'. 5.1 and Fig.2. 'inputd 3' and 'inputd 7'.APPENDIX D TYPICAL INPUT FILES The file. are for comparing Huh's And De Vahl Davis and Mallinson's correction formulas in the problem geometry. 4. Fig.
0 0.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 0.0 0.4 0.01 0 1.0 0.0 2 l.0 1.0 1.045 0.120 1.0 1.0 0.0 0.0 10.0 0.0 60.5 0.0 0.0 2.0 0.2 10.0 0.0 0.t20 1.0 1.0 0 0.0 4 0.0 2 5 10.0 I 3 I 16G 1.0 0.0 1.0 10.5 6.0 0.013 0.0 0.75 10.0 1.0 1.0 1.4 5 1 1.0 0.0 0.44 0.0 4.4 0.0 0 2 1 3 O0 2 2 O O ADI 10.0 0.4 1.0 10.01 0.0 0.0 0.7 0.0 0.0 1.09375 0.0 1.0 1.0 0.0 60.0 .0 0.0 60.0 0.0 1.9 0.0 0 0 1.0 0.0 0.0 0.
0 NATURAL CONVECTION IN A CARTESIAN CLOSED COMPARTMENT 1. 20 1.2 1.751.5 0.0 4.20 0.67 1 1I 2.FILE: INPUTN DAT A VM/SP CONVERSATIONAL MONITOR SYSTEM 'PAGE 001 5 10 0 1 0.0 1.0 0.0 0.20 2 1.0 0.20 030 1.0 0.01 0.0 2.21795 1.0 I 43 1.7 1.0 1.44 0.0 0.0 1.0 1.0 8 2 I 3 1 3 0 3 0 1 1.0 32.0 68.045 0.5 5 1.0 6 6 50.90 0.0 0.5 1.0 1.0 10.+20 60.20 1.0 0.0 0 0 2 1.0 0.0 1.0 0 1.0 0.0 1.20 II I 1.0 0.0 60.0 1.0 0.20 1.4 0 0 0 0 O 0 0 6 2 0.0 677.01 1.0 1.0 .
07107 12 12 Ill 2 5.009 G 1.9 0. 020 2.0 9.0 1.014 ).948 1.0 0.207 0.07107 0.999 1.497 2.0 1.911 3 4 1.0 1.0 1.0 .0 0.07107 0.497 0.0 9.0 0.997 7 1.07107 111 3 1 1 2 5.0 1.0 0.003 7 1.0 0.0 0.0 1.OGI 2.0 8 10.0 9 10.0 9.0 0.0 1 .000 1.5 1.009 4 1.0 7.0 0.0 0.0 GO.063 0.0 10.0 1.0 0.000 0.0 0.0 7.026 5 1.07107 0.07107 1.0 0.01 0 0.0 0.07107 7.063 ( ).0 0 0 NUMERICAL blFFUSION .0 0.0 12 11 2 121 t 5.5 0.47.0 7.0 0.3 1 I 1.70711 0.000 0.940 1.0 0.0 0.0 0.01 0.0 1.0 5.014 0.000 10 1.0 0.0 0 2 5 3 0 7 6 1.045 1.0 1.0 1.09375 0.0 4 I 36 1.0 1.013 2.511 1 1.207 0.0 5.000 0 0 O O I.0 0.05.0 0.0 9.0 60.0 0.751 2 1.75 0.204 0.000024 0.0 1.0 0.II I 11 2 2 1.+20 I 0.0 2 1t 1 113 5.0 0.0 7.44 4.0 0.001 0 1.002 0.0 0.457 1.0 9.0 0.991 6 1.000 0 0 0 O .CASE 3 2.0 1.0 1.O.457 1.0 0.0 0.0 5.0 0.0 0.0 0.0 0.249 3 1.0 9.0 1.000024 60.0 0.0 0.974 5 1.000 10 1.O 2 1.0 0.FILE: INPUTD 3 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 10 10 0 0.5 1.002 0.0 2 4 2 4 1.70711 1.204 0.0 1.0 0.0 0.000 9 1.0 1.0 7.061 1 0.0 1.409 1.0 2 1.02 1.0 1.
457 2.0 4 1.0 0.0 1.0 1.001 0.0 7.000 6 1.00.9 0.143 0.0 0.5 1.75 2.0 4 2 4 1.0 2 9.633 1.907 3 1.0 0.0 0.0 3 0.17051 0.00.4 8 2 1 3 0 0.0 5.0 60.3 1 1 1.07107 1 1 713 2 5.0 0.626 1.000 0 0 0 1.0 0.0 1.0 I 1.0 0.045 1.333327 60.0 3.0 NUMER ICAL DIFFUSION .167 1.003 17051 1.0 0.990 5 1.0 10.0.01 0.0 1.0 0.000 0 O 0 0 0 7 1 2 7 2 5.47.0 0.0 0.09375 .911 9.0 2 7 8 5.0 1.07107 711 8 8 2 5.0 7.01 2.0 0.001 0.0 .0 1.0 1.0 1.+20 2.0 0.009 1.0 1.07 07 0.44 4.0 0.0 2 2 0.0 5.0 0.013 2.000 6 1.4 1.0 1.0 0.0 7.002 5 1.014 0.071107 0.0 10.0 0.07107 I 113 7 2 1.0 0.003 0.457 2.07107 7.0 1.0 1 9.01 0 0.0 1.FILE: INPUTO 7 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 6 0.0 1.0 9.0 1.0 0.0 4 I 36 1.0 0.0 7.0 0.0 0 0 2 2 1 0.633 1.0 0.0 0.5 1.013 4 1.CASE 7 0.0 60.0 7.0 0.374 1.0 0.0 0.0 0.0 1.0 0.07107 0.420 1.0 1.0 0.0 0.143 0.0 0.014 0.0 5.0 1.0 2 0.333327 1.
L Z01 'L L'09C91C P09c91''? 1'Q9C9p 0'19C91' 1 iLO'O 1'LO*0 GL00O LLO'0 £00.*I riI 00 0*0 0*0 0i'00 0 10*0 6 v 0*0 0'0 top' .SAS 3~~~)Vd kOIINOW 0 0'0 0*0 0.z 10'o 0 1 0'l GL'O 1O1 '0 I 0*0 01l 0'l L910* 199*0 099.0 01l 0'00' C0'I C. 1'0 0.0 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 ~W31.0 0.0 GLO10 ZU10 SL9*0 109.0 CGO. 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 .09 0.0 09G"0 9Cs£0 661''0 6sv1'. 0 1'0l '0 011 '0 sCl '0 LSI *0 6LI 0 G0Z 0 cpz *0 01 01 0 01 oz 61 LI 01 01 91 cI zI 01 6 0 Ia z 0 6ZC *0 OLC 0 &Pt'v O CZi..0 9?LL '0 G'.u *0 Ul' I 6P I 0011* I 0.09 W '09 0. *0 PI 9.NOIS(fJJIG IV:)IUIW(IN too 39vd IVNOIIVSU3ANIX) dS/W~A v VIIi III)II OIMINI O~N :311A II . 0 LG0 .1bl b 690*6 £00*0 LC9 0 90c.1 1' 0*0 0 *1 D99C00 0'1 0IO' 0*0 0*0 0*0 0'0 0 *I0*0 9c 1 t.
0 26 26 5.0 0 2 1.20 2 1.20 2 1.20 911 1.FILE: INPUTO IIU4U VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 2 0.0 8.0 2 25 25 5.20 I13 1.20 9 1 1.20 913 1.6603 0.0 0.20 1 1.0 .0 0.0 0.0 2 25 5.20 1011 1.0 5.6603 0.0 I 1 5.20 10 1.0 5.6603 0.20 5.
007 0.013 2.072 ' II 12 13 14 15 16 17 10 19 20 21 22 23 24 0 0 0 I '.030 0.034 0.0 1.0 0.0 1.003 0.0 60.223 3 1.120 2.0 2 2 2 0.02.0 1.0 1.0 0.0 1.076 0.0 5.0 5 1.566 0.009 0.0 1.4 0.0 1.0036G0 0.0 1.0 0.rILE: INPUIO OM A VM/SP CONIVEIlSATIONAL MONI(fro SYSTEM PAGE 001 24 f( 0 I 0.01 0 0.5 1.636 0.0 0.2406 1.005 0.0 1.GOG 2 1.0 1.496 0.197 0.OGI 0.522 0.0 0.360 0. 4G1IG04 7.003660 60.002 0.567 9.099 0.021 0.0 0.3 1.0 0.0 0.951 0.0 0.002 2.0 0.4GI6 0.0 0.0 0.767 0.705 0.5 1.0 0.932 1.0 0 2 1 3 0 7 6 I 0.0 1.02.029 0.144 1.0411 0.0 0.420 0.003 0.0 1.093 0.463604 2.026 0.0 1.02 2.246 0.0 0. 1.33G 6 1.969 0.463604 2.302 0.0 1.174 1..0 0.011" 0.44 4.5 1.526 3 4 5 G 2.9 0.0 0.0 60.260 0.0 0.002 0.045 1.0 1.0 0.0 I 1.0 1.0 1.0 0.0 0.0 5.20 1.0 0.051 0..0 1.0 0.0 3 10 1.156 0.463704 9.0 0.30i9 1 0.0 0.0 1.920 0.0 1.09375 0.0 0 0 NUMEtICAL IrFFUSION ..0 0.0 1.133 10.0 0.030 0.903 0.75 1.OM 0.0.0 0.365 9.0 1.242 7 1.460 4 1.000 0.4 4 1 36 1.0 1.0267 0.403604 2.0 1.044 9.0 0.96G 0.115 0.949 0.463G04 2.0 1.700 .0 0.0 0.016 0.0 2 4 2 4 1.0 1.0 1.01 0.003 0.0 0.0 1.
: INP.20 I 2 1.0 0.0 .20 9 1 1.0 0.20 913 1.20 911 I.0 26 2 1.0 I 5. & .0 0.20 5.0 2 5.0 5.0 0 29 25 I 25 2 1.0 I*  .SILr.20 1011 1. 6603 0.0 2 5.10 OM A VM/SP CONVERSATIONAL MONITOnr SYSTEM PAGE 002 2 0.0 5.20 1. GG03 0.6603 0.0 26 5.20 113 1.