This action might not be possible to undo. Are you sure you want to continue?
Kang Yul Huh and Michael W. Golay Energy Laboratory Report No. MITEL83011 August 1983
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS
by
Kang Yul and
Huh
Michael W. Golav
Energy Laboratory and Department of Nuclear Engineering Massachusetts Institute of Technology Cambridge, MA 02139
MITEL83011
August
1983
Sponsored by: Boston Edison Co. Duke Power Co. Northeast Utilities Service Corp. Public Service Electric and Gas Co. of New Jersey
2
ABSTRACT
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS by Kang Y. Huh and Michael W. Golay
A computer code is developed to predict the behavior of the hydrogen gas in the containment after a lossofThe conservation equations for the four coolant accident. components, i.e., air, hydrogen, steam and water, are set up and solved numerically by decoupling the continuity and momentum equations from the energy, mass diffusion and turThe homogeneous mixture form is used bulence equations. for the momentum and energy equations and the steam and liquid droplets are assumed to be in the saturation state. There are two diffusion processes, molecular and turMolebulent, which should be modelled in different ways. cular diffusion is modelled by Wilke's formula for the multicomponent gas diffusion, where the diffusion constants are dependent on the relative concentrations. Turbulent diffusion is basically modelled by the ke model with modifications for low Reynolds number flow effects. Numerical diffusion is eliminated by a corrective scheme which is based on accurate prediction of crossflow diffusion. The corrective scheme in a fully explicit treatment is both conservative and stable, therefore can be used in long transient calculations. The corrective scheme allows relatively large mesh sizes without introducing the false diffusion and the time step size of the same order of magnitude as the Courant limit may be used.
A
3
ACKNOWLEDGEMENTS
This research was conducted under the sponsorship of Boston Edison Co., Duke Power Co., Northeast Utilities
Service Corp. and Public Service Electric and Gas Co. of New Jersey. support. The authors gratefully acknowledge this thank all those who assisted them
The authors
during this research especially Dr. Vincent P. Manno, Mr. B.K. Riggs, Ms. Rachel Morton and Prof. Andrei Schor. work was expertly typed by Mr. Ted Brand and Ms. Hakala. Eva The
1 Truncation Error Diffusion in a One Dimensional Problem Another Approach for a One Dimensional 4.2 Limitations 2.3.4 ADI Schemes Conservation 4.1 4.1 Mesh Point Implementation Mesh Interface Implementation 4.2.2 Turbuletit Diffusion 3.3.1 Assumptions 2.1 Continuity and Mass Diffusion Equations 2.3 Basic Assumptions and Limitations 2.1 Skew Differencing by Huh 4.1.3.1 Derivation of Turbulence Equations 3.2.2.3.3.2 Physical Constraint 4.1 Problem Statement 2.4 4.3.4.4.2 .2 ke Model 3.4 Solution Scheme 2.3 Low Reynolds Number Flow CHAPTER 4: NUMERICAL DIFFUSION Truncation Error Diffusion 4.2.2 4.4.2.2 Truncation Error Diffusion 4.1 4.4.3.1 Molecular Diffusion 3.4.3 Tensor Viscosity Method Corrective Scheme by Huh 4.2 Comparison of Explicit.2 Crossflow Diffusion 4.1.2 Governing Equations 2.4. Implicit and 4.1.2 Energy Equation 2.1 Diffusion Convection 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 2.3.2.5 SMAC Scheme and Compressibility CHAPTER 3: DIFFUSION MODELLING 3.3 Truncation Error Diffusion in a Two Dimensional Problem 4.3 Discussion of Skew Differencing and Corrective Schemes Raithby's Scheme 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
3(B) . . . .. . .15 . (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 . Donor cell solutions with various diffusion corrections for a pure convection problem with 8 = 450 . . 142 5.. 137 5. . CA in Fig.convection problem with 6 = 450 Comparison of the true solution and solutions by donor cell scheme.3(B) . 5.10 Comparison of the numerical solution with donor cell differencing of convection term and analytic solution with increased diffusion constant. .430 . . .. . Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 10x10 meshes along the line CA in Fig..14 140 5. . 135 5.8 .43* .. . 144 . . True solution and donor cell solutions with and without diffusion correction for a pure convection ..11 Comparison of the true solution and solutions by donor cell scheme.12 . . . skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure . . problem with 6 = 63.3(B) . . . . numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig. 5. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with 6 = 63.16 143 5. 5. . . . ... . . .9 5. 136 5. . .13 139 5..17 Comparison of the analytic solution. by De Vahl Davis and Mallinson's formula along the line BD in Fig. 5. ..
. ... .. ....147 5. 5. . ... . ...22 150 5. . .. . . . 5.10 5. .. .. donor cell solution and solutions by the mesh point and mesh interface implementations of the corrective scheme for the recirculating flow field given in pig..21 S. Simple recirculating flow field for test of the implementation strategies of the corrective scheme .. 146 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 . . ... . 152 5.. 5. 153 . .. . . .. Diffusion constant corrections at each interface for the mesh point and mesh interface implementations of the corrective scheme for recirculating flow field given in Fig. . . . .. .19 . numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig.19 . .18 Comparison of the analytic solution. . .. . . . . 151 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 . 145 5. .. ..20 . . 148 5.. ...24 .3(B) . Comparison of the true solution.23 . . ..19 .
30 for At = 0. . Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. . . . 0. .30 5. .. 5. 5.. . with donor cell differencing of convection term without any diffusion correction . .. 1 178 . Steam concentration distribution after one time step. . 156 5. . .. . . . . . . . corrective scheme . . . . with mesh interface implementation of the . . . . .28 .. 0.. . 157 159 5.30 for At = 1. .29 Steam concentration distribution after one time step. . ...cx T where 4 is given by. . Problem geometry for comparison of the explicit and ADI schemes .0 sec and the Courant limit of 1.32 161 5.31 1i60 5.1I 5. . 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 . 5. . ..0 sec .30 with .0 sec .5 sec and the Courant limit of 1. .27 . 154 5. Steam concentration distribution after one time step. 155 5. . .. . . . different time step sizes Boundary conditions in terms of 4 for the problem in Fig. . 0.3 sec..0 sec along the line AB in Fig. . 5. . .3(B) = e. ... .3 sec.26 Diffusion constant corrections of the mesh interface implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .. .3 sec. .33 Comparison of the ADI solutions for the profile of 4 at t = 10.. . 163 A..
. . . . Page . . .12 LIST OF TABLES Number 4. . 78 4. . .114 . .2 . . . . . . 2D. 1981 . . . donor cell differencing of the convection term . . . . . . . Comparison of the Von Neumann analysis and numerical experiments for the stability condition in terms of the cell Reynolds number in an explicit. . .1 List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research. . .
D y.13 NOMENCLATURE Letter A c c Definition Area Specific heat at constant pressure Specific heat at constant volume Courant numbers in x and y directions (1) Dimension (2) Diffusion constant (3) Divergence (4) Discriminant of a quadratic equation Binary mixture diffusion constant between air and hydrogen Binary mixture diffusion constant between air and steam Binary mixture diffusion constant between hydrogen and steam Crossflow diffusion constants in x.C x y D D ah D as Dhs D .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 .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.
tan.1 v u Mesh configuration. 1 tan STime a Z Superscript (1) Vorticity (2) Amplification factor in Von Neunann analysis constant of heat transfer Traction tensor due to viscosity Summation Time step n .14 Letter R Re Re Rt Rx S Sc T t At u v v w Ax y Ay Az Greek Definition Universal gas constant Reynolds number Real Turbulence Reynolds number Reynolds number in x direction Source Schmidt number Temperature (1) Time (2) Parameter of a quadratic equation in Von Neumann analysis Time step size xdirection velocity ydirection velocity Velocity vector zdirection velocity xdirection mesh spacing Mole fraction in multicomponent gas ydirection mesh spacing zdirection mesh spacing a ala2 8 6ij C p v 4(1) 6 61 Thermal diffusivity or diffusion constant Parameters used in Von Neumann analysis Thermal volumetric expansion coefficient Kronecker delta Turbulent dissipation rate Density Kinematic viscosity Vpiscosity Phase change (2) Any general conserved quantity Velocity direction.
15 Subscript cell cf DM f g i=l 2 3 4 L t m mix new ob old R t total US Cell Crossflow diffusion De Vahl Davis and Mallinson's Fluid Gas Air Hydrogen Steam Liquid Droplets Left hand side Liquid Molecular Multicomponent gas mixture New time step Obstacle Old time step Right hand side Turbulent Total calculation domain Upstream .
steam and liquid droplets are assumed to be in thermodynamic equilibrium and the relative humidity is assumed to be 100%. The four components. The research work reported here is primarily con cerned with the formulation and validation of the physical models and numerical schemes in the second slow mixing stage. hydrogen. The governing conservation equations are decoupled in order to simplify the solution procedure. The error due to decoupling is negligible in a slow transient where the . air. From the hydrogen transport point of view.__ 16 CHAPTER 1 INTRODUCTION One of the major concerns in the Three Mile Island (TMI) accident was hydrogen gas accumulation in the containment. The distinct feature of the second stage is a much longer time scale in comparison with the first blowdown stage. the fluid dynamic phenomena in the containment in order to justify those remedies. the response of the containment during an accident can be divided into two stages.I~~U1i* 11*11044~1 1~.( 1~111111_~. Some simplifying assumptions are made concerning the thermodynamic state in the slow mixing stage. Some remedies have been proposed to deal with However. it is necessary to understand such problems. the fast blowdown stage and the slow mixing stage. although it may be less than 100% when there is no liquid component present.
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 turbulent diffusion constant is predicted by the kc model [45).17 state change over one time step is small. The molecular diffusion constant is predicted by Wilke's formula [73) for multicomponent diffusion and ChapmanEnskog formula [5) for binary diffusion. and the total diffusion constant is the sum of the diffusion constants of those two mechanisms. In the first step the continuity and momentum equations are decoupled from the energy and other scalar transport equations and solved by the Simplified Marker and Cell (SMAC) method in order to obtain the flow field. The turbulent kinematic viscosity. k. Finally. resulting in the same convection velocity for the four components. Diffusion occurs by two independent mechanisms. Convection is assumed to occur as a homogeneous mixture. Convection and diffusion are the central issues in physical modelling efforts of hydrogen transport. are determined by their own transport equations. c. The turbulent kinetic energy. the phase change is taken into consideration to maintain 100% relative humidity. The diffusion constant. of each component in multicomponent gas depends on the mole fraction of that component. molecular and turbulent. and turbulent dissipation rate.
Truncation error diffusion is a one dimensional profile error and crossflow diffusion is a multidimensional operator error [68) of the finite difference equation. The turbulent Prandtl and Schmidt numbers are assumed to be equal to one. the term. There has been much debate on the numerical diffusion and many schemes have been suggested for the past two . The effective diffusion constants of the two errors are of the same order of magnitude. 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 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. 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. Therefore. numerical diffusion.18 which is the diffusion constant for momentum transport can be calculated directly from k and E. Since the error usually appears as an additional diffusion. The numerical diffusion has two different sources. Truncation error diffusion occurs in the flow direction while crossflow diffusion occurs primarily in the direction normal to the flow.
19 decades. This situation is partly due to misunder standing of the sources of numerical diffusion and also partly due to use of inappropriate approaches for its elimination. 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. although with different stability conditions. However. The conservative property is essential in a long transient problem like the hydrogen transport in the containment. Chang's methods [16) are examples of skew differencing schemes and Huh's corrective scheme and tensor viscosity method [25] are examples of corrective schemes. Raithby's [53) and S. mesh point and mesh interface implementations. ADI and implicit schemes. The mesh interface . The schemes presently being used can be divided into two categories. have been tested for recirculating flow problems. The corrective scheme can be implemented with any of the explicit. skew differencing and corrective schemes. The corrective scheme is inherently better than the skew differencing scheme in that it is conservative and does not affect the simple solution procedure. Two implementation strategies for the corrective scheme. 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.
simulation results compared with experimental data. [49) includes all the Development Laboratory) [6].20 implementation has always given physically reasonable solutions and may be used extensively for all diffusionconvection problems. . The suggested physical models and numerical schemes have been used to simulate the LOCA experiments performed in BattelleFrankfurt [44) and HEDL (Hanford Engineering Ref.
21 CHAPTER 2 PROBLEM STATEMENT AND SOLUTION SCHEME 2. for dealing with this problem. In order to justify the design of any mitigation system. the hydrogen has received much attention in the safety analysis of nuclear power plants. use of flame suppressants and enhanced venting capability. There have been some mitigation procedures suggested.. The major safety concern is that of keeping the containment pressure below a certain level to prevent a large scale leakage of radioactive materials. After a while the slow mixing stage follows the initial blowdown stage and continues for an extended period of time. e.g. Thereafter. hydrogen generation gave a serious concern about the integrity of the containment in the TMI2 accident. . In addition to the pressure increase due to the primary coolant. it is essential to understand the fluid dynamic phenomena in the containment. installation of ignition devices.. a large amount of steam and water will come into the containment in the fast blowdown stage increasing the containment pressure. 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.1 Problem Statement Once a lossofcoolant accident (LOCA) occurs in a Light Water Reactor (LWR). containment inerting.
thermal nonequilibrium. etc. Therefore.. phase change and heat transfer between the gas components and wall.g. e. some simplifying assumptions should be made to use the numerical procedure without impairing the acceptable solution accuracy. laminar and turbulent flows. 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._ __ _ __ _ I_ ~ 22 Since a numerical method is suitable for this purpose.
T) (2.5) Turbulence dissipation rate: D 1 _ 1t StE xk Dt P Txk 0E ] + C t*( 1 p k ui xk ouk au.1) Momentum: pgI + d + V0 (2.3) Mas s diffusion: S+ V*vp i .4) Turbulence kinetic energy: Dk 1 t 3k 3 P 1 k ok x .6) Equation of state: p = f p (ii .xk k 2 v t BT C2 + kg k Pr 3x 2k k t (2. (2. I+ 3x 1 . Continuity: V'v= 0 pl + Vvv ] = Vp + Vv = 0 (2.t = VD.Vp i + 4.7) .2 Governing Equations The governing conservation equations are set up to describe the postLOCA fluid dynamics in the containment. The physical implications of the conservation equations will be given with their basic assumptions and limitations.S + + t Bui P au k aui + ax~ ) x k X axk gk Pr t t 'T xk (2.2) Energy: p[e e + Vvh] + = VkVT + e fg (2.23 2.
9) D. .18) p = Epi (2.19) Note that 1.T) (2. = D.T) (2.13) D. The subscript i denotes the four components as .8) f (v) (2.17) t = C k 2/C E (2.24 Constitutive equations: 'N = f (p .11) k = . Im + Dit it (2.k + k t m Dim = fD(Pi'T) (2.12) (2.10) v = vm + Vt (2.15) Dt = v t (2.14) m = fk (PiT) (2.16) kt = . = f (Pi.t (2.
2. The mass diffusion equations are summed up for the four components. Therefore.1 Continuity and Mass Diffusion Equations The continuity equation is redundant with the four mass diffusion equations and there is'an inconsistency between them. u. (2. . H2' steam. p. four mass diffusion and two turbulence equations with ten primary unknowns.22) There are eleven conservation equations. The phase change occurs between steam and liquid droplets and the diffusion constant of liquid droplets is equal to zero.2. w.k C for a three dimensional case. liquid droplet. 3. T. v.21) D4 = 0 (2. one continuity. one energy. The turbulence equations are written in terms of the Cartesian components in tensor notation. air. i=14. there is one moreequation than is required. three momentum.i (4).25 follows: i=l i=2 i=3 i=4 2.20) i = 2 = 0 43 + 04 = 0 (2.
2.23) The sum of the phase change terms is equal to zero. Since the concern is only for the slow mixing stage after a LOCA.26 are satisfied.Vp. 2.26) It is a reasonable assumption in a turbulent flow regime that the four components have the same diffusion constant.1 and Eq. ap at S+ Vp = E(VDiVp i ) 1 (2. 2.27 remain valid throughout the transient.) + V(Ep Jt i = (V'D. 2. 2.25) Vp i = 0 (2.) i + E 1 (2. 2. at p= 0 (2. 2.26 can be reduced to the following Eq. .25 and Eq. it is reasonable to assume that Eq.(Ep i ) a .27.4 are consistent only if the temporal and spatial variations of the total density are very small. Vp = 0 (2.26 .24) Now it can be seen that the continuity equation is valid if and only if the following Eq. Eq. 2.27) Consequently Eq. Therefore.25 and Eq.
is also of an approximate nature and the exact form is given in the following.iei) + V'[Ep. E it p + e.2 Energy Equation The energy conservation equation. Be. ae. It is clear that the major contribution to this term will come from the latent heat of the phase change. (p. 2. (2. .ih i ] = V*kVT (2.30 is substituted in Eq.2.i + V. 2.27 2.31 so that the internal energy and enthalpy can be written in terms of the temperature. = VkVT (2.2.31) The thermal equilibrium assumption is incorporated in Eq.3. 2. Eq. 2.30) Eq.29) The second term on the left hand side of Eq. LeaPi LeQ.29 denotes the energy change due to the concentration change in a given control volume.i = V'kVT (2. p + £e.28) The temporal term can be divided as follows. I at t + V [vp.29 to give the following.ih. [c pihi .
T 1 Eq.T P P e Piei p ih p.32) where ei i = eg4g + e 4) Sefg g e.32 become identical by defining average internal energy and enthalpy as follows. It is necessary to clarify these assumptions and their limitations for a safe use of the given physical models and solution scheme.28 (Picvi )~ +V"[vTZpic pi] + efg 9kVT = (2.33) h (2.34) 2.3 Basic Assumptions and Limitations The governing equations are based on some simplifying assumptions about the physical phenomena in the containment after a LOCA. 2.c .1 = c Vl. . P (2 33) (2.T h. = c p1 .3 and Eq. Pic vi T c p p. 2.
Dt and k t over one time step should be small because the turbulence equations are decoupled from the velocity field calculations. .3.1 Assumptions 1.2 Limitations 1.3.29 2. The temporal and spatial variations of the total density should be negligibly small because of the incompressibility assumption. 4. The turbulent Prandtl and Schmidt numbers are equal to one. with no slip. 2. 2. The four components are treated as being nearly incompressible and the phase change rate is moderately small. 3. 5. vt. The relative humidity in the containment is 100%. The change of the diffusion constants. 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. 2. The phase change rate should be moderately small because the solution scheme decouples the phase change term from the energy conservation equation. The four components are in a state of thermodynamic equilibrium. 3.
the conservation equations are decoupled into two sets.7 are a coupled set of equations with primary unknowns.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 SMAC scheme is In the second step. p. k.5. In the first step the SMAC scheme is used to get the convergent velocity field with zero divergence for every mesh.7. which are the four mass diffusion equations. . Pi(4). 2. 2.1Eq. is used to update the reference pressure and reference density which is important in calculating the buoyancy force.30 2. T. E. the obtained velocity field is substituted in the scalar transport equations. 2. The equation of state. the relative humidity may be less than 100%.i Since it is too difficult to solve the whole system of the equations and obtain a consistent solution for all the unknowns. It solves the . When there is no liquid component left. continuity/momentum equations set and other scalar transport equations set. 2. Eq. explained in detail in section 2. v. . energy equation and two turbulence equations. u. In the third step phase change is taken into consideration to maintain 100% relative humidity.4 Solution Scheme Eq. The phase change is assumed to be zero in this step.
37) Py The vorticity conservation equation is derived from Eq. 2. Continuity: au ax av ay 0 (2. the divergence of the velocity field vanishes everywhere.36) .1_ _~ 31 continuity and momentum equations to get the velocity and pressure field in an iterative way. Since the continuity equation used in the SMAC is an incompressible form. which can arise with net inflow or outflow boundary conditions..ax 2. the SMAC scheme can be used only for incompressible flow calculations. However.36) ydirection momentum: av av ua av 1 + pg + a y x2 Nv 2 (2.35) xdirection momentum: au au + u+v at ax au ay 1 p ax + Pgx + a 2u 2 ax 2 2au + 2 2 (2.37)y (2. 2. it is possible to modify the SMAC scheme to accommodate a slight compressibility effect.37 as follows.36 and Eq.37): . (Eq. i.e. 2. The continuity and momentum equations are given in two dimensional Cartesian coordinates in the following. ay (Eq.
36 and Eq. 2. 2. 2.37 as follows.40) S (2.39) where av au D= .+ y ax ay The basic idea of the SMAC scheme is to separate each calculational cycle into two parts. (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 (2.32 a C 2 a2 + u+ V = v+ ) at . 2. ay 2.37 as follows. the socalled tilde phase and pressure iteration.38) =ay ax The Poisson equation for the pressure field is also derived from Eq. ax 2 2.36) + ~(Eq.x ay ax 2 ay2 where au av (2. .41) where fn and gn are the explicit quantities at time step n.36 and Eq.
the following relations hold after the tilde phase.40 and 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. shows that the vorticity is independent of the pressure field. a guessed pressure field or the pressure field at time step n is used to start the iteration. 2.42) av .v) D (2. _ Pi+j2P +P iPij+ pij f(uv)  Dn + l D n D D Ax 2y t (2.44 is given in the following. The impor tant point is that un + l and vn+l calculated from Eq. 2.44) Eq. 2. The equation of vorticity transport. au ay au av ax  (2. V2p = f(u. Eq.40 and Eq.39 may be rewritten as A finite difference form of Eq. follows.43) ax ay It can be seen from Eq.33 Since the pressure field in Eq. 2. 2. Therefore.45) .38.41 have the right vorticity. 2. 2.O 0 (2. 2.41 are still unknown.
Once the zero divergence velocity field is obtained. (6u) L =  pAx (6u) R = t6p pAx (2. sidering the cell (i. equal to zero. while the continuity equation has the following compressible form.47) ) The pressure correction is given in terms of the residual divergence for every mesh. 2..49) .46) (2.48) The SMAC scheme can be extended to a slightly compressible fluid flow. The velocity correction formula is derived from Eq. The slight compressibility means that the NavierStokes equations may be solved safely with the incompressibility assumption. 2.37 so that it does not change the vorticity implemented in the tilde phase. S+ at V"(p) = 0 (2.j) 26p( 1 Abx 6p = h + _) Ay 1 At( A + 1 Ay 1 Dt D (2. no more correction will be made.36 and Eq.34 The pressure correction formula is obtained by conn+l only and putting D.
Now the problem is how to converge to a predetermined nonzero divergence field. V: total volume of the containment.p_ p at 1 min/V At n p n n (2.uniform over the whole containment. : in n =Vp v  total net inflow of mass at the boundary during the time At. This can be done with the following modification of the pressure correction formula of the SMAC scheme. .35 From Eq.50) The two terms on the right hand side of Eq. The associated physical assumption in Eq. 2. 1 . 2.50 may be given from the previous time information and boundary conditions. 2.52 is that the pressure perturbation due to the inflow at the boundary propagates throughout the containment instantaneously and the pressure and density increase is. 2.49 the nonzero divergence may be derived as follows.52) (2.51) v p where m.51 and Eq. D = V'v 1 p p at V p (2.
48.50. 2.6p = . The velocity correction formula remains the same as Eq.53) where D O is given by Eq. .(DD 1 1 + ) 2At( Ax Ay (2. 2.
Convection occurs as a homogeneous mixture resulting in the same convection velocity for the four components. hydrogen. molecular and turbulent.37 CHAPTER 3 DIFFUSION MODELLING The hydrogen. the process Since there may be three gas of diffusion is also increased.1. The thermal conductivity and viscosity are also diffusion constants in a broader sense for momentum and energy transport. It gives the diffusion constant of each component in terms of . air and steam. As collisions occur more frequently. Although turbulent diffusion is greater than molecular diffusion by a few orders of magnitude. steam and liquid components are transported by convection and diffusion in the containment. the diffusion constant of each component is calculated separately by Wi~k's formula [73] in Eq. proper modelling of the latter is important because the molecular diffusion of hydrogen is significantly greater than those of other gases and also because there may be a laminar flow region in the containment. air.1 Molecular Diffusion Molecular diffusion occurs by the collisions of gas molecules. in the containment after a LOCA. 3. Diffusion occurs by two different mechanisms. components. which should be modelled independently. 3.
1Y A B DAB 1Y B yA DBA 1Y C YA CA C YC CB B YC DBC C D = C YB DCB (3. stants are given as follows..2.: Binary mixture diffusion constant between the ith and jth components.0018583 1 + M A MA MB (3. T3 D = 0.2) poAB D.: yi: Diffusion constant of the ith component Mole fraction of the ith component D. Eq. The property of air will be replaced The resulting diffusion con with that of nitrogen gas.AB where DAB: AB Binary mixture diffusion constant in [cm /sec] .38 the binary mixture diffusion constants and mole fraction of that component. The binary mixture diffusion constants are calculated by the ChapmanEnskog formula.1) where D. 3.
4) np = np 2 where the correction factor 4 is given by.3. In order to calculate the para meters aAB and 0DAB in Eq. P n ]. 3. the following relations are required.lll~*. 3.5) 2 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. The error range of the ChapmanEnskog formula is reported to be about 610% [5).AB: Dimensionless function of the temperature and intermolecular potential field for one molecule of A and one of B.. (3. ap = np £ (o n + o p) EE e (3.3) CAB = AB For the pair of polar and nonpolar gases a correction factor is introduced to account for the polarity. For nonpolar gases OAB and ODAB are calcu lated by Eq. =n [ p n = [+ S n . OAB = (o A + oB (3.2.
40 The parameters in Eq. H2 a n = 7. 3.65 A The polarizability is given for the hydrogen and nitrogen (air) components.9492 x 10 D ah D 1 T .9 x 10 25 25 25 [cm3 3 [cm ] : N2 : an = 17.9035 = 4.6 x 10 The fitting functions for the binary mixture diffusion constants are obtained by the leastsquarefit method.= 1. The containment pressure is assumed to be1 atm. t* .6947 as . 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 are defined as follows.6810 x 10 .2 E/K = 380 0 K a = 2.6 T1. 5 = 4.
= n i=l y.10) Viscosity in [poise) Temperature in [K]) LennardJones parameter in [A) Dimensionless nTumber .8916 x 10 where T: D: [OK1 [cm 2 /sec] 5 T 1.8144 (3.lllI_ 41 Dhs = 2.8) j=l where ij 8 Mj j ij M The viscosity of nonpolar gases may be obtained by the following [5).k.7) k.67 x 10 where : T: a: Svi : o2 2 (3.lii.__Li~n_~PIIII ~.6) The viscosity and thermal conductivity can be considered as diffusion constants for momentum and energy transport and calculated similarly as follows [5). = 2. n mix = Yipi n i E j=l Yj ij (3. i (3.
3. 200 C There is another formula suggested for the viscosity of multicomponent gas in the following [5]. The ks model sets up the transport equations with some empirical constants for the turbulent kinetic energy.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. c (3. There are several models suggested for Presently the best model calculating turbulence effects. The effective diffusion constant is therefore the sum of the molecular and turbulent diffusion constants. The .y RT k=l PMiDik k i There is a useful relationship nonpolar polyatomic gases.11) for the Prandtl number of Pr = c (3. E. k. 2 n Yi umix = C 2 i=l Yi + 1.385 nn y. and turbulent dissipation rate. seems to be the kc model originally developed by Launder and Spalding [45).009 cp at 1 atm.[5].42 The viscosity of steam is. Wsteam = 0.
= u. (3.' For example. etc. 3.' etc. u. 34. + u.1 Derivation of Turbulence Equations [15.= v ( Dt + ui j u P= Dt ax x 2 k6 ij 3x_) 3 i (3. 45] The transport equations for the turbulent kinetic energy and turbulent dissipation rate are derived from the continuity.u.' j T'= a t x. momentum and energy equations by decomposing the variables into mean and fluctuating parts and averaging the resulting equations.2.14) . u.u.15) . + Pi Pi = Pi + p. 29.43 turbulent viscosity is given directly in terms of k and C. . The effect of turbulence appears as the additional terms due to the product of fluctuations which do not necessarily cancel out.13) (3. These terms are treated by the eddy viscosity concept of Boussinesq and eddy diffusivity concept which are given in the following.
3. 3.2 ke Model The transport equations for the turbulent kinetic energy and turbulent dissipation rate form the basis of the kE model.16) . t + at r a (ruk) + (wk) = [ p r (r oka r ) + (t ) az ok a au aw + ) r aw au 2t + 2( + 1[2() 3 3r p + 2 + u2 2 ] .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. 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. j=i.' u is transformed to the turbulent kinetic energy equation by the contraction. A two dimensional cylindrical coordinate form of the kE model is introduced in Eq.16.44 The transport equation for u.2.
17) The suggested values of the constants are given in the following table. it is difficult to determine whether a region under consideration is in a turbulent or laminar flow.3 The molecular effects can be included in Eq.09 C1 1.44 C2 1.0 o 1. C 0. Since the kE model is applicable to fully . 3.45 The turbulent viscosity pt is given in terms of k and E in Eq. = C pk 2/ (3. 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.metry. 3. 3.3 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.92 ok 1.17.16 as follows.
19) = pk2/VE .0. 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. Eq. 3.lit k + l ak aui i + t (ax k + xk auk axi axk e .3 exp(R2t ) (3.5/(l+Rt/50)] C2 =2 where R 2 0 [1. The following kc model is a modified form by Jones and Launder [40.2vak 2() k 2 (3.18) In the above equations C 1 1 ok and a retain the values assigned in the ordinary kc model. CP = CP0 exp[2. 41). D Dt p ax k [(t Dt+ ) e k + C k i + k x ) au axk C 2 E2 2 t 2. it is required to extend the model to laminar flow regime or to switch to laminar models according to some criterion.16. The model presently being used is just to use the sum of the molecular and turbulent contributions for every region in the containment.46 turbulent flow.0 a ui ( 2 p ax ax 2 ) Dk Dt 1 p ak [(.0 . while C1 and C2 are to vary with turbulence Reynolds number. Rt.
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.8) 3 @n + u()n + a( ax i+b ax 4 n+g _ n )n+g . n+g 4 i+ Ax n i Ax 1 xji+ n i+l . 1 < c < n 2 4 + i Ax 2 11 4 2 1 i+d < d < 1 The mean value theorem.y+b) = f(x.a(?) ) 2(.x+a) and [y.3.51 Subtracing Eq. we obtain the error at time step (n+l). 4.6 from Eq. . n+l i+f EXACT n+l i FD n i+f  4P n a n+g i + At [ u( x ) i+a (4. Eq. It can be extended to a two dimensional form as follows.7. 4.i+c T i+d ax where the following relations have already been used. af f(x+a. 4.a i = ( nn+g n+g ax i+a  < a < n i1 8a ( ) n i+b a 2 4n+g ax 2 i+c 2 ox n 1 < b < 0 n+g (a n+g ax i. n .y+b].y) + (f) af (4. is for a one dimensional case.
F (4.10 are at some appropriate points in the domain [xi . 4.52 Now Eq. 4. The terms A.10) The derivatives with respect to time and space in Eq. 4. 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.9 reduces Eq.8 to the following form. 2 n+l (4i ) EXACT  (4+) FD = i n+1 f )2 2fx 1f 2 ) (f 2 2 .D +aAx(cd)Atax 2 +agt .C ax 22 +At g t . . they may represent the errors over many time steps in a stable scheme. D and F are time derivative terms which become negligible for slow transients.10 are based on the assumption of perfect information at time step n. 4.tax .A fAxAt axat 2 B utx(a+b) . Although the error terms in Eq. C and E are truncation error diffusion terms while the terms B.xi+ ] and [tn tn+). E .
1 Truncation Error Diffusion in a One Dimensional Problem The truncation error diffusion in a one dimensional problem is analyzed for a steadystate case with no source. 4= Cecx + C2 where c = u/a . The analytical solution of Eq. u ax a22 (4.1.12) and C 1 =N cL 1 e C2 ecL 1 . o 1 2 3 L i1 1 i+l N 4(x=0) = 40 N Q(x=L) = NAx = L Two boundary values 0 and QN are given as constant.53 4.11 is given as. cL (4.11) ax Consider a case in which the domain is divided into N equal meshes and the mesh size is Ax. The results may be extended to a general one dimensional problem if the effects of the transient and source terms are not dominant in determining the profile of 4. 4.
4.11 is finitely differenced using donor cell scheme for the convection term in the following.14 may be recast to Eq.16) .16 is easily obtained from Eq.i + 4i2 Ax = a (4. i+l i = (P+l)( 1 .15) Equation 4. 4. . P(4). Equation 4.14 is the cell Reynolds number or cell Peclet number according to whether the diffusion constant a is the kinematic viscosity or thermal diffusivity. where P =u a  l i24 + ii (4.13) where the velocity u is positive. i u x i+l2. i+  i = (P+l)(4)  4)i (4. Equation 4.15.13 can also be solved with the given boundary conditions by reducing it to the following form. (4.15 and solved for in subsequent procedures.40) .54 Equation 4. 4.14) The parameter P in Eq.
2 1 = (1+P) (4 0) 4' N 92 = N2 (1+P) 2 (  0) = N.41 = N ) (1+P) [(1+P) N1  2) (4. 4.17) 01 can be expressed in terms of 40 and 4N N N by Eq. 4. 4.1 (1+P) N 1 (4.20 have been obtained without any approximation for Eq.LI II .13.I_. Both the analytical and numerical . we obtain the solution for 4. i  0 = 4.~LI LIII~. 4.19 into Eq. 4.17.12 and numerical solution Eq. as.~l*IIIIX1~~^ ~ ^I ^~Xi 55 Then. (1 0 ) is ) P( N ¢0 (1+P) given as follows.18) (l+P) + 1 Therefore.20) Now the analytical solution Eq.1 Inserting Eq.19) (1 . 4. (1+P) 1 .(1+P) P (1+P) 0 + N . 1 (1+P) i . 4.16.'N1 (1+P) N1  ) 0 N .11 and Eq.
The in analytical solution Eq.13. + 24 S i i+1 i1 =e e l NP  1 (e P P 1) 2) ) ( (e 0 (421) The finite difference solution Eq. e (il)Pe (ePe1)2 NPe .24) It is also possible to go through the similar procedures with a central differencing form of the convection term in Eq. 4.25.23) (4. Eq.1 e Therefore. i+l 2Ax il S S2ax 4) i+l 24i + i 2 i (4.13.12 is substituted for 4i Eq.+1 i + ) il 0 (4. 4. and the effective diffusion constant for the numerical solution will be obtained.20 is also substituted for 4.13. 24 j. 4.1 (4. 2 ii P ( (1+P) (1+) N _ 1 N 4). Pe = in (1+P) (l+P) iIp2 N (1+P) . 4.21 and Eq. 4.22) Comparing Eq.25) . 4.56 solutions will be reinserted to the finite difference equation. in Eq.25) (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. 4.
1. it helps understanding the origin of the truncation error diffusion.57 The effective Peclet number for the finite difference solution of Eq. 4.2 Another Approach For a One Dimensional Truncation Error Diffusion The truncation error diffusion in a one dimensional problem can be evaluated in a different approach. 4.26 may be a good indication of the truncation error diffusion occurring in a general one dimensional problem. Eq. 4. the transient and source terms and the mixed type of boundary conditions will not affect the numerical diffusion appreciably if they are not dominant terms in determining the profile of 4. Although this approach is approximate and overpredicts the truncation error diffusion.26) where 2 < P < 2 The effective diffusion constant De can be readily obtained from the following relation. Pe = kn (2P) 2+P (4.27) In general. Therefore. uLx D De Pe  (4. 4.24 and Eq. The convection terms in . As in the previous section the velocity u is assumed to be positive and constant.25 can be derived as follows.
29 from Eq. Pe also goes to zero resulting in no diffusion error in Eq.( i+f 2 (4. i. 4.28.58 Eq. the diffusion is dominant over the convection. the truncation error diffusion constant for a convection dominant problem may be given approximately as follows. D uAX 2 (4. 4. the profile of 4 will be linear and the truncation error will be reduced to zero. 4. 4. i. ae u( x where 0 <f < ae + u() ax i+  u~x a2 ) .e.31) ND ..26.4 and Eq. the convection is dominant over the diffusion.28 and Eq.24 and Eq. It can be shown that as P goes to zero.5 may be approximated as follows. u i Ax i1 u i+ a4 u() ax u(a) ax (4. 4. 4.28) Ax (4.30) Both Eq.29 hold only when the Peclet number is large. 4. If the Peclet number is small. Therefore.29) i+ The truncation error diffusion can be quantified by subtracting Eq. 4.e.
2 u + v 2 + +( ) (4.1.33) . conservation equation with no source term is given in the following. 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. Since the gradient of 4 in the flow direction is negligible in a convection dominant problem.59 4. 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. u i 6x ilj + v i by ij_a(i+1 2 4x i + il1 +ij+l 2i 2 + 0iji by (4. The formulas for the truncation error and crossflow diffusion constants derived in this chapter reveal that they are approximatly of the same order of magnitude.32) = x 3y Equation 4. A steadystate.3 Truncation Error Diffusion in a Two Dimensional Problem There are two sources of numerical diffusion in a two dimensional problem.32 is finitely differenced using donor cell scheme for the convection term. two dimensional.
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 .35) . then the truncation error of the convection term is given as follows.1.v( ay ) ij .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 = .x[cos + (cose tane 1 + sine) + sine tane ] ax xY (4.60 The approach in section 4.2 is applied to Eq.j+ + v(a) ly i+h.n) ( = x cos6 + y sine I = x sine + y cose (4. transformed to the coordinate ( .33. 4.y) is by rotation.
y v e Sx Then Eq. e is equal to 0 or 7/2. 4.35.36) where U = /u2v 2 Now Eq. 4.n) given by Eq.32 is reduced to the following form.61 where 5 is the coordinate in the flow direction.34 will be transformed to the coordinate system (t. 4. dominant truncation error diffusion occurs When only in the flow direction. 2 (A)  [(sine tane1 + cose) 2 + a2 (sine + cosetan6 1 ) a + (0) 4 an U" (4.37 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. the problem is basically one dimensional and the truncation . 2 2 2 + a = ( 2 ) (4.37) Equation 4.
which shows a single mesh with pure convection.1(A) is transformed into . The origin of the crossflow diffusion is illustrated in Fig. In Fig. 4.1. 4. The crossflow diffusion comes from the multidimensionality of the problem. that will be named here as the crossflow diffusion. the hot and cold fluid enters the mesh from the left and bottom surfaces at 45* angle. The hot fluid will come out of the top surface and the cold fluid out of the right surface.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 4.68) and it was clarified that this is the dominant source of the error in most multidimensional problems. In the donor cell scheme Fig. 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. 4. It is entirely different Recently this addi from the truncation error diffusion. tional false diffusion was explained by Patankar [52] and Stubley et al.l1(A). [67.3.62 error diffusion occurs in the flow direction as in a one dimensional problem. therefore it exists only in two or three dimensional problems when the flow direction is not aligned with the mesh configuration.
Illustration of the crossflow diffusion in a single mesh with pure convection .1.63 Hot Mledium Hot 7 Hot Cold (A) Cold (B) Cold (C) Fig. 4.
in The numerical results Fig. 4.64 Fig.1(B) where the velocity components normal to the surface are considered. diffusion occurs in multidimensional donor cell differencing of the convection term. 4. 4. 4. A 2 . This is a pure convection problem Consider the fluid element The fluid with no physical diffusion.1(B) are again interpreted as Fig. . In the donor cell scheme it is distributed all . and the sum of the values This distribution along the points A1 . . ] = (lp)4 = where u . at those points is exactly equal to 4. 4. The value of 0 at the point A should reappear at the point X because there is no physical diffusion.1(C) by a Therefore. CAB which looks like a long onedimensional rod. . Another illustration of the crossflow diffusion is given in Fig. an appreciable amount of false program user.and intermediate temperature fluid will come out of the top and right surfaces. A 3 .3. .&x u v ax Ay The effective diffusion constant for the crossflow diffusion will be calculated in the two dimensional Cartesian coordinate in Fig. 2 (lp)[l+p+p + . of the value of 4 causes the crossflow diffusion.2 which shows a row of meshes aligned in the xdirection. Then homogeneous mixing occurs in the mesh.
Illustration of the crossflow diffusion in a row of meshes aligned in the xdirection with pure convection . 4.2.65 (1p)4' x p (1p) 2 p (lp) p3 (lp)4 p4 (1p) A5 Ay A Ax 2 4 y p t* U where p = Ax Ay Fig.
V A2 e A.3. Geometry for explanation of the crossflow diffusion in a two dimensional Cartesian coordinate .66 4. 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.
right and left hand sides. The travel of the fluid element is analyzed in Fig.t 0 PA £3 (Current) 2 1 constant ) where 1 = (Gradient) = U3 Lx/cose. In the donor cell scheme it is divided into two portions. 4. the following right hand side diffusion constant results. The diffusion occurs both to the The diffusion to the right hand side is considered first in the following. p4A and (lp)4A and appears at the points Al and A 2 . which shows that the value 4A diffuses out along the fluid element. Gradient) A /k 1 =A of 4 Current 2 2U of 4 Therefore.4.67 element was on the line CAB at time t in the velocity direction. k2 = Ax sine/sin(el+8) k3 = Ax sine 1 /sin(e 1 +8) After substituting appropriate expression for each term. and is translated The value 4A at the point A should reappear at the Point A'. v. Diffusion A t1 . sine cose cos= sx Uacos61 sin(O+e ) I DRHSS .
.ct t2 t3 Fig. 4.68 ti . Geometry for explanation of the crossflow diffusion in a rotated two dimensional Cartesian coordinate .4. Ya.
82 x 9x 2 2 By 2 2 The current on the one dimensional fluid element can be factored into the x and y components. sine cos6 D = UAx sin cos8 x sin(e+e 1 ) 1osel 2 sine cose sin 81 D = UAx y sin (8+8 ) cos1 1 (4.69 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. D D = uAx(lp) = VAy p (4.38 is the effective diffusion constant when the diffusion is confined to the one dimensional fluid element.39) Equation 4.38) Equation 4.40) . sine cose RHS LHS cos1 sin(+8) (4. The crossflow diffusion constants are the ratios of the current and gradient of 4 in the x and y directions. There are two diffusion components in a two dimensional problem as follows.39 is identical to the following simple expressions.
that is to say.6 the value at the point 0 is 4 and that value should reappear at the point X where the velocity vector meets the plane ABC.42 underestimates the crossflow diffusion as shown in Fig. (4. 4. D DDM = UAlxy sin2e= 3 3 4(by sin36 + Ax cos36) UAx tan6 1 sin26 3 4(tan8 sin6 3 + cos3) (4. Eq.42 give the same result. RHS D = It can be shown that. Now the analysis will be extended to a three dimensional case in Fig. De Vahl Davis and Mallinson [24] also derived the effective crossflow diffusion constant and their result is given in Eq.42) Equation 4. In Fig. < 6 < 81 or < . 4. 4. When both 6 and 61 are equal to .40 and Eq. 4.70 where p = u Ax u v By Equation 4. 4.6. 3 0 when tan 6 = tan6 1 .41 shows that DRH S (or DLH S ) is maximum at an Tr7LHS angle of 61 < 6 e between 4 and e1.41) Equation 4.42. 4. In the donor cell p' y scheme the value 4 is divided into three portions px 0 .5.38 should be differentiated with respect to 6 in order to find out the velocity direction with maximum crossflow diffusion.
0 0 0 .5 I I 0.0t Dcf UAx de Vahl Davis & Mallinson Huh 81=8C 1. 0. Comparison of the prediction formulas for crossflow diffusion constant by De Vahl Davis and Mallinson and Huh .5.5 80 \ % % 4 1.71 Cross flow Diffusion 2.0 150 300 450 600 750 900 Fig. 4.
4. Geometry for explanation of the crossflow diffusion in a three dimensional Cartesian coordinate .6. p Fig.72 z C V Az X 0 o a X AY Ay B y x z C p .
CX which are denoted as £i' z2' k3 can be obtained as follows.73 z occurring at the points A. This is the socalled crossflow diffusion phenoThe currents due to the crossflow diffusion are . menon. 4. cosa cosB cosy (4. X(£ 4cosa. £4 cosy) nB .43) Since the point X is on the plane ABC. B.44) 4 Therefore. and C instead of at the point X. BX. n C denote The coordinate of the Any point on the plane ABC should satisfy the following Eq.43. t4 cos8. point X is given as follows.and nA unit vectors along those directions. XB. S= 4 + y z zAx 1 cos+ + cos Ay Ax + cosy Az (4. S+ Ly Ax + Az = 1 plane ABC (4.45) The lengths of AX. along the directions XA. the following relation holds. XC. £1 = AX =/(x4cosa)2 + (4cos)2 4cos) 2 = Ax 2 + 42 2Ax£ 4 cosa .
48) The currents in the directions nA . n C can be obtained in terms of their components in x. (Ay4 cosS).47) etc.46) + The unit vectors nA . Z 4 cos8. nB' n C can simply be given as the products of the transported quantities and velocities. 4 cosy] 1 nB = [i4cosa. z directions. Therefore. nB.  n A = XA +1A  OA . 2 o C 14cosy] = 1  £ 4 cosa.OX (4. (Az 4 cosy)] (4. I (Current)xAi XAI = px Px C1 1 Lt . y.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.
(Grad)x = 4. y.50) Now the x direction crossflow diffusion constant may be given as the ratio of the current to the gradient of 4 in the x direction. Px D (Ax£ + Py D + nC(4 (Current)x 4 cosa) c osa) At ( 4 (cosa) (Grad)x = 4/Ax Therefore.51) p Ax D x = lt (Axk 4 cosa) + p Ax .49) The gradients of 4 in the x./ Ay 4/6x Q/by (Grad)z = 4/Azl (4.(4 pz x co s a ) ct 4 + at ( 4 cosa) 4 .75 I(Current) XB = Py p2 t I(Current) XCI where At = k /U z '3 At (4./ Ax (Grad) = 4. z directions are given as follows. (4.
1 = uAx(lpx) (4.53) The crossflow diffusion constants in the y and z directions can be obtained in the same way. although none of them has been successful enough to be accepted widely. p Ax D = ux[ £4cos x . Dy = vLy(lpy) Dz = wAz(lpz ) The rotation of (4. finite element method.54) 4. indices will also give the same result. we can simplify the result as follows.52) Inserting the expression for At given by At = £4 /U. .76 p D x Ax At A2 At t cosa(p +p +p) 4 x y z SxA At Ax L cosa at 4 (4.3 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. etc. This section is a review of the schemes that The have been used to eliminate the numerical diffusion. tensor viscosity method.
. skew differencing and corrective schemes..1 can be categorized into two types. S. The corrective scheme is to reduce the diffusion constants in order to compensate for the additional false . In the skew differencing type schemes finite differencing of the convection term is done not in terms of the x and y directions.3. but just appreciably lower than that of the donor cell scheme. In the skew differencing scheme the 5point relationship becomes a 9point relationship.^ILli. Since the upstream point in the velocity direc tion does not necessarily fall on the mesh point where the quantity under consideration is defined.II1~LI1I~~ ICII1_1___~1___*_____11__ 77 Most of the schemes in Table 4. the interpolation should be done between neighboring points to obtain the value at the upstream point. The donor cell treatment of the convection term results in a 5point relationship in a two dimensional problem. but directly along the velocity direction.4 fall in the category of a corrective scheme. Basically the inclusion of four additional corner points contributes tomore accurate numerical modelling of the convection. the interpolation between neighboring points may still give some crossflow diffusion. 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. However... The crossflow diffusion is not eliminated entirely in the skew differencing scheme.~_ 1 i_..
inte Conribo:r/me.Highe cdtr upwi:nd :"'ie Cdi't.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.thod LLxW'en doff  21 x11 80 x 400 Table 4..78 Chss' iction Cen:ered . 1981 [65) .j./LLUE Huint/QUICK Wilkes/LUE Lillimpon/SUD LiiL. Grandotto Hickmott Firs.1.ordc.tnce Vuer:o: upwind Yes Yes Yes Yes 2) x II 21 x ]1 21 x 11 21 x 11 41 x 21 41 x 21 21 x 11 81 x 41 28 x 14 Se Tzb)t 2 Up'nd ft(e elemmnt !.ronLUD Lillinion/VUDCC Nuana.1QUICK t ETovh ci .aptjve me.e elements Clifft Dontcl t al.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 . £Ebzhau t aL/PATANKAR Lilinpion Orlandi Priddin Wada cit l. viscosity method Sel. Wilkes Elbahu ct l. List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research. upwind f~iite di Teretn atl.hod Mesh *txLnsformation rr.thod of chaaceriics Fiite anuryic method Tenso. Esposito Glas and Rodi Huffenus Lnd Khali.
54.53 and Eq. 4. the corrective scheme can give a numerical solution which is almost free from numerical diffusion.54. The tensor viscosity method is one example of the corrective scheme. D. 4. and D z in Eq. 4. The finite difference equation for Raithby's scheme is derived on the assumption of a linear profile of 0 normal to the flow and a uniform profile in the flow direction.55) where n is the distance normal to the flow and U = /u 2 +v 2 . y and z directions will be reduced by the amounts of D .79 diffusion. where the balance equation is set up for a given control volume by considering convection and diffusion at the control surface. 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 = C1 + C2 n = C1 + C2 (  (4.53 and Eq. The corrective scheme is valid only if the effective numerical diffusion constant can be predicted accurately.3. 4. although the validity of its correction formula is not clear. 4. The correction formula The diffusion con used here is Eq. It is therefore a conservative scheme.1 Raithby's Scheme Raithby's scheme [53] is an Eulerian type of approach.. stants in x.
7. the calculation domain showing grid lines.80 Fig. control volume and notations in Raithby's scheme . 4. A segment of.
2 Skew Differencing By Huh Skew differencing scheme here treats the convection term in a Lagrangian way. Although this scheme is nonconservative.3.57) where the upstream value €US may be calculated by .= U ax dy SC (4.56) where U = /u is the coordinate in the velocity direction and 2 +v 2 . This should be repeated for the four control surfaces of every mesh in a two dimensional problem. 4. 'ij US (4. u 4 + 'v. which may be too complicated and time consuming for a practical purpose. The inclusion of the four additional points also complicates the structure of the coefficient matrix in an implicit scheme. The convection term is differ enced in the velocity direction between mesh points.81 The convection quantity at the control surface is calculated from the upstream value in the flow direction and the upstream value is obtained by interpolating two neighboring points. it is much simpler than Raithby's scheme and becomes conservative in a unidirectional flow. not in terms of the convection that occurs at mesh interfaces.
uv (4. DD y . the tensor viscosity [25) T = Lt uu u2 The expression for is given in the following.53 and Eq. z directions to compenx sate for the crossflow diffusion effect. 4. D are derived . D . y. 4. DD z in x. The validity of this scheme is based on the fact that the crossflow diffusion can be predicted accurately. 4. Some of the results of this skew differencing scheme is given in section 5. the crossflow diffusion constants D in section 4.54.82 interpolating two neighboring points.58 for the numerical diffusion constant is apparently not consistent with the results in Eq.3. Dx = uAx(lpx) D Yy = Ay (lpy ) The expressions for .2.3. DDx.2. 4.3 Tensor Viscosity Method The tensor viscosity method treats the numerical dif fusion constant as a tensor quantity.4 Corrective Scheme By Huh The corrective scheme is to use the reduced diffusion constants. 4.58) t2U= Eq.3.
which is used with the cell dimensions to get the crossflow diffusion constants. Dy ._IIU__lpULI~~I___~lLlt i~Lli ~i ~iL~pn~sl. Two implementation strategies have been tested so far. 83 D z = wAz(lp z ) where px = u (4. 4.59) v u . After the crossflow diffusion constants are assigned to every mesh point. the mesh point implementation and mesh interface implementation.Dz at the center of a mesh where all quantities are defined except the flow field.+ .1 Mesh Point Implementation The mesh point implementation is to assign the crossflow diffusion constants Dx .+ w Ay Lz x etc. The velocities at two boundary faces are averaged to get the representative velocity.3. .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.. they should be averaged again between neighboring mesh points to calculate the crossflow diffusion current at mesh interfaces.4. In a coarse mesh or recirculating flow field the latter implementation strategy is definitely preferred.
8. U and V. For the second case the two outward velocities. 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. U1 and U2.84 4. Dyl and Dy 2 . The two velocity sets (Ul. For the first two cases the cross flow diffusion constants are zero because the velocity vector is aligned with the mesh configuration.Vl) and (U2. as shown in Fig. and the resulting D x and D y will be assigned to each interface.4. 4. The velocity U is split into two components. The third case is expected to have a flow split as shown in the figure. so that they are proportional to V1 and V2. can be directly used with cell dimensions in Eq. The same logic can easily Appendix C includes be extended to a three dimensional case.3.59. . The mesh interface implementation is always recommended although it has a more complex program logic than the mesh point implementation. The right interface will have the crossflow diffusion constant (Dx1+Dx2 ) and the top and bottom interfaces.V2) are used to give (DxlDyl) and(Dx 2 'Dy2 ). There are three possible cases that should be treated independently in a two dimensional case.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. 4.
Three possible flow configurations for mesh interface implementation of the corrective scheme in a two dimensional problem .8. 4._~I~~ _~~~~_I__~__~~ __ 85 (1) (2) (3) Ul U2 V2 VI U U V14V2 VI+V2 V2 U2 = UV2 U = Ul + U2 TV2 Fig.
2.86 the fortran program for calculating crossflow diffusion constants in an arbitrary flow field. 1. 4. 4. 4.4 Comparison of Explicit.) = i1 n D Ax ( i+1 2i l n n + n ) + iS i1 (4.60) Explicit (lD): n+l I n i At + u (i i x n i. Implicit and ADP schemes There are three typical solution schemes. The pros and cons of the three schemes will be compared in the following four viewpoints. 3. for the general conservation equation. Eq.60. Implicit and ADI (Alternate Direction Implicit).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 ) (4.62) . Explicit. conservation physical constraint accuracy stability 2 ax 2 2 y 2 +ux +at + 7 + T7) + S sz (4. which is a parabolic partial differential equation.
when the net inflow or outflow at .87 ADI (2D): n+ ij (A) _n ij In+ + u ij _ cn+ in + v n ij1 At 2 Ax Ay )n+ i+lj 24 n+ Ax 2 +n+ n l + D j+l n ij 2 Ay n+l ij n iji + S n+l ij n+ ij n+ ) ij n+ i1 n+l ij1 At 2 Ax Ay n+ =D i+i S2. 2 Ax n+l 2 * ij v n+l ) ij n+l Ay 13j1 n+1 ij+l n+l 1j1 At Ay2 + S (4. + 4).63) 4..4. or affected only by the source term contribution (SO).1 Conservation The conservation principle is that the total amount of 4 should be conserved (S=0). 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.
the conservative property may not be crucial. where the convection terms of momentum equations are linearized in a nonconservative way.4. there are some nonconservative solution schemes like ICE (Implicit Continuous Eulerian).88 the domain boundary is equal to zero. Although they usually give . the conservative property is as important as the stability because the error from nonconservation may accumulate over many time steps. All the explicit.2 Physical Constraint The physical constraint is closely related with the stability in a practical sense. + CI 4. however. Although a conservative form is preferred to a nonconservative one. For a long time transient. the same exchange term should be used for neighboring meshes. 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. In order to guarantee conservation. implicit and ADI schemes can be made conservative by using appropriate differencing forms for the convection term.
The laws of physics impose some restrictions on the numerical scheme to get physically reasonable solutions.D .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. )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.4.1 Diffusion There is a maximum net flow that can occur between neighboring meshes by diffusion.2. new values of 4 are given as follows. 4. R C AxAy = CAxy + D R Ax Ax CAy C y At At ¢4AxAy = 1RAxAy . they are from entirely different origins.
65 may be a too conservative criterion in a real problem. scheme given in Eq. R AxAy(4) C) > 2D (R Ax CyAt 2 DAt X2 < 1  (4. Both of the oscillation and instaThe ADI bility should be avoided in a transient problem.65) Ax Equation 4. we have the following physical constraint. When the physical constraint is ignored and the time step size greater than that given by Eq. 4.64) Eq.65 is used. t' > Q' .64 or 4. should still hold. . the solution will be unstable or stable with damping oscillations. Therefore the physical constraint should be respected in addition to the stability condition to get a meaningful transient solution. d 2D t < 1 (4.90 Since the relationship. however. 4. 4. there will be a damping oscillation in the solution when the time step size is greater than that given by Eq.64 or 4. 4.65.63 is unconditionally stable.64 may be extended to a ddimensional case as follows.
B 4y S The explicit and donor cell treatment of the convection term gives the convection quantity at interface S as ulAAtAy. the damping oscillation or instability may occur as in the case of diffusion. than the Courant limit. Then uAttay is no more an appropriate expression for the convection quantity during At.66) When a time step size greater than the Courant limit is used.91 4. 4.4. Therefore. At < 'x u (4. It is based on the assumption that 4A is the repreIf At is greater sentative value of the control volume A. a portion of the control volume C will cross the interface S.2.2 Convection C A u B C A ).3 Accuracy Realistic interpretation of a finite difference equa tion comes from an integral form of the conservation .4.
4.92 n+l n+l 4' I I. I I t i r _ I I [ n n+1 n n+l _ 0 It Explicit Implicit CrankNicholson $i+.9. 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.
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. The stability can be checked by the theorems about matrix eigenvalues or Von Neumann analysis. In other words. the error propagation is suppressed and the largest error contribution will be from the previous time step. the accuracy depends on the profiles 4 in time and space because the exact profiles assumed for are not known a priori.93 equation. accuracy depends on how well we can compute the convection and diffusion quantities at control surfaces over the time At.4.4 Stability The stability is a mathematical problem. 4. 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. It depends on the maximum eigenvalue of the iteration matrix. The accuracy expressed in terms of O(At). The infinitesimal approximation of differential terms is not appropriate in most problems of our consideration. The Von Neumann . 0(Ax 2 ) is The also not appropriate for finitely large At and Ax. If its absolute magnitude is less than one. whether the error introduced at a certain step will increase or decrease as the calculation goes over to the following steps.
1 One Dimensional Central Differencing of Convection Term The finite difference equation is given for a one dimensional general conservation equation with central differencing of the convection term.4. n+l i n €i+i I + u i+1 n ii 2Ax i_1 lt n n i+l . is expanded in Fourier series as follows. (4. lowing sections. 4. 1 4n = E~nei where I = .1.2 2i + D i+1 i n 2 i i1 (4. stability criteria will be derived for the explicit.68) .1 Explicit Scheme The Von Neumann stability analysis of a finite difference equation is simple in its idea.. implicit and ADI schemes using the Von Neumann analysis. 4.4.94 analysis is to expand the solution of a finite difference equation in Fourier series and check the decay of each mode separately.4.4. The stability is not a sufficient conIn the fol dition for an acceptable numerical scheme. but its algebra may get complicated.67) The solution 4.
the function . mode to achieve stability.70 may be rewritten as follows by defining cos6 x as t. 2 11 = [1 .72) In order to get the stability of Eq. the following expression for = lCxsin xI where cx d x uAt A tx  2d (lcos x) (4.70) Equation 4.67. in Eq.95 The absolute value of should be less than one for every When Eq. 4.67.2dx(cosex) 2 + [Cxsinex 2 < 1 (4. f(t) = i 12 = (4d2_ 2 )t2 + (Sd 2+4d )t + C 2+4d 24d (4.2dx(t) where t = cose 1 < t < 1 ] 2 +C 2(1t 2) 1 (4. [i .71) The function f(t) will be defined as follows. Therefore. 4.68 is substituted results. 4.69) Dbt Dt 2 Tx.
In case (1) f(t) is a parabola concave upward and f(l) should be less than or equal to zero as shown in Fig.1]. f(t) < 0 (4.96 f(t) should be less than or equal to zero for any t in [1. Therefore. f(1) = 0 (4.73 can be solved by a graphical method using the characteristics of a quadratic equation.2d )2 > 0 (4.73) for any t in [1. The parabola f(t) always meets the taxis because the discriminant is always greater than or equal to zero. D(Discriminant) = (C 2 . 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.74) One of the two meeting points with the taxis is the point t=l because f(l) is equal to zero.75) The parabola f(t) may have three distinct shapes according to the sign of the coefficient of the second order term.10(1).1] Equation 4.
 r I I t i t 1 1 t 0.. according to the sign of the coefficient of second order term . 4. f(t).97 f(t) f (t) f (t) 'I I 4.10..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. Three distinct shapes of the parabola.
4.98 f(l) = 8dx(2dx1) < 0 0 < d < (4.78 is a finite difference equation with donor cell differencing of the convection term.dx) is shown in Fig.4.C (4.11.76) In case (2) f(t) is a straight line and f(l) again should be less than or equal to zero. 4d 2 . t=l as shown in Fig. 4. 4.76 and Eq.2 One Dimensional Donor Cell Differencing of Convection Term The following Eq. 4d 2 C 2 4d C > 0 < 0 : : 0 < d dx > < 2 Cx The stable region in the coordinate system (Cx. we get the following stability conditions. 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.1.4. .77) Summing up the results in Eq.2d Axis of symmetry: t = 4dx dx > C 2 x > . 4. 4. 4.10(3).77.
4. Stability ondition in the plane (Cx.d ) for an explicit scheme with central differencing of convection term in a one dimensional problem .11.99 d x 1 1 0 1 dx = C2 x 2x C Fig.
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)
4.13. The region in the complex plane where (a +c ) should exist for the stability of an explicit conscheme with donor cell differencing of vection term in a general two dimensional problem .104 Im \ .1 SRe 2 al +2 1+a +a 2 Fig. .
The region in the complex plane where the amplificatior factor. Fig.Re Both al and a 2 should be in the shaded circle to ensure that Ial+a 2 +1I< i.105 . .14. . should exist for the stability of an explicit scheme with donor convection term in cell differencing of a general two dimensional problem .. 4.
89 can be solved by the graphical method as in the previous sections.4. 4. our usual concern is can be checked directly.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 .4 Stability Condition in Terms of Cell Reynolds Number The stability conditions in the previous three sections are in terms of the dimensionless numbers. 4.106 Both Eq. Therefore. and time step.1.e Rx .e Y) (4.dx) and (C y. At. A three dimensional case can also be solved through the same procedures and the results are given in Fig. which is independent of the time step size. the stable time step size for a given mesh spacing.88 and 4. 4. 4. Cx and d .d) should be in the shaded region for stability. the stable range of Cx should be determined in terms of the cell Reynolds number Cx/dx'. the stability However.4.88 and Eq. At.16.15. The stability condition is again given by Eq. 4. 4. Ax.90) a 2 = C [1 y R R Y + R (4. When Ax and At are already given. The stable region shrinks on a linear scale with dimensionality. which depend on both the mesh spacing. where the points (Cx.91) . The results are given in Fig.
4.idition in the planes (Cx.dx) and (Cy.107 d (d ) d = x 2 1 x + 1 4 C x(C ) 16 d= C2_ C x 2x x Fig.dy) fcr an explicit scheme with donor cell differencing of convection term in a general two dimensional problem.15. Stability cc. .
dx) etc. for an explicit scheme with donor cell differconvection term in a general three encing of dimensional problem .16.108 d xy(d or d z ) Cx d dx + 1 1 Cx (C d! S3 2 . Stability condition in the planes (Cx. 4.(C Cx x y or Cz) z d ) Fig.
95 ) The function f(t) satisfies Eq. [ .94 ) D(Discriminant) = C 2 (K2C ) 2 > 0 (4.88 and Eq. 4.90. 4. 4.92) Equation 4.C (+2x . (1) Cx 2 K2 1) > 0 concave upward . f(t) = C (1+K) 2 (lt) .92 can be simplified as follows.94 and Eq.__LI___IIII__LI*X__ ID I 109 The following condition for Cx and Rx can be derived from Eq.93) ) where 2 K = 1 +.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. 4.1] f(l) = 0 (4.95. and has distinct shapes for the Following three cases.) RX (1cose )) x + [C sine] x x < (4.C K(lt) + C ( l_1t = (C 22 KC 2 2 )t2 + (2C 2 22 22 K +C K)t + C 22 K  C K + C 2 < 0 (4.
therefore the subscript x may be replaced with the subscript y in Fig. V f . 4.110 (2) C 2(K 21) = 0 linear (3) 22 C 2(K 2_1) < 0 concave downward For case (1) and (2).97) The results in Eq.17. f(l) should be less than or equal For case (3) the axis of symmetry of the parabola The should be on the right hand side of the point. 4.96) R 1 x < 0 < C x < (1+2) Rx . (4.96 and Eq. t=l. The stability condition in Fig.97 are shown in Fig. 4. although it can be relaxed if there is a constraint on the velocity direction and mesh spacings such that the following' relation is satisfied. 4. R > 0 x : 0 < C < x 2 2 2(1+) x and (4.17 is applicable to all general 2D problems.98) Ax . solution procedure is similar to those in the previous sections and the results are given in the following as. to zero.yu < 1 (4.17. 4. The same result will be obtained for the ydirection.
Stability condition in the plane (Rx. 4.Cy) for an explicit scheme with donor cell ditferencing of convection term in a general two dimensional problem .17.111 x (C y C x  2 1 2 (1+ L) R x C = 2(1+ ) x 2 0 R (R x y Fig.Cx) and (Ry.
112 1 1 x CX x 1 (l+f) (1+ x ) 1+f 1 1 f 0 x V where f =y u Ax Fig.Cx) for an explicit scheme with donor cell differencing of convection term in a two dimensional problem with f v u/Lx .18. Stability condition in the plane (Rx. 4.
uLx.17 should be Table 4. 4.C Y y . 4. the stability condition is found to be the following. 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 (4.20.102) dx > .2 Implicit Scheme The following Eq.99) (4.2 shows the com 1 l+f in Fig.101) (4.4.113 Then the number replaced with on the C axis in Fig.19 and Fig.17 and numerical experiments. 4. and it can be seen that the implicit scheme is unconditionally stable if the diffusion constant is than . 4.100) For a general two dimensional problem.18. 4. 4.4.99) The stability condition is shown in Fig. d >  greater Cx (4.99 is a fully implicit finite difference equation for a general one dimensional problem and the Von Neumann analysis will be applied in the same way as in the explicit scheme. (4. parison between the analytical result in Fig.Cx d > . 4.
120 0.120 0.629 2.660 0.3 8.333 833.403 0.147 0.660 0.780 0.019 Numerical Experiments 0.488 0.333 4.833 8. R x 0.540 0.260 0.114 Maximum Stable C.499 0.380 0.333 83.2.060 Table 4.167 2.720 0.180 0. Comparison of the stability conditions in terms of the cell Reynolds number by Von Neumann analysis and numerical experiments in a two dimensional explicit scheme with convection donor cell differencing of term .083 Von Neumann Analysi s 0.
4.dx) for an implicit scheme with donor cell differencing of convection term in a one dimensional problem .19.115 d =  x d =. Stability cndition in the plane (Cx.C x 2x x 2x C x dx 1(C2+C 2xx ) Fig.
Stability condition in the plane (Rx.20. 4.116 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 .
63.104) .103) jsj <1 In order for I. 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.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. 4.4.+d )] C ) {[1+(lcosex +dx)) +[fsinex }{[l+(1cosy) (+dy) + Y[siney } (4. C d x>2 (4. 2 _C C 2 {[(cos )(+d X))] C [sine I2} {(1cos6 C 2 )(. the following condition should be satisfied.4.63 gives the following stability condition. One of the formulations has both the x and y direction components in each step. The Von Neumann analysis of the first formulation in Eq. 4.117 4.
106) C d y  >  ~ 2 (4. C d y>  2  4 (4. . 4.63 can also be given by the Von Neumann analysis as follows.63 has the same stability condition as the fully implicit scheme. 4.107) Therefore the second formulation of the ADI scheme in Eq. C d x > 2 2x (4.118 The same condition should hold for the ydirection.105) The stability condition of the second formulation in Eq.
1 Natural Convection Results The containment is modelled as a two dimensional rectangular compartment with an obstacle as a heat sink. Natural convection occurs due to the heat transfer between the obstacle and containment air. Finally. 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. the ADI scheme is explored to use a time step size larger than the Courant limit to save the overall computational time. 5. Updating of the reference state. density is determined by temperaSince the natural convection of air ture and pressure.2. For example. 5. energy convection and heat transfer models are considered and the resulting flow fields are given in Fig.1 and Fig.1 Updating of the Reference State The state of air is determined by any two independent properties. 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. 5.
8 sec .8537x10tu/ft20 F sec (Cp) = 1.4 sec AT = 100F T .29 Ibm A= 2 ft2 T.1.0 Btu/lbm 0F (Q ob = 208.5 ft/sec h = 1. 5. = 680F air T = 580F Fig. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 10. 0. = 677.120 p ).
0 Btu/lbm OF (c )b = 208. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 62.2. = 677.5 sec .4 sec 1 0 AT = 10 F T air = 680F .29 ibm (K 2 A= 2 ft T.8537x10 tu/ft2 F sec p)ob = 1. T ab = 580 F Fig.121 > 0. 5. 5 ft/sec h = 1.
Told' Tnew: Temperature of the fluid at old and new time steps. time Pressure of the fluid at old and new steps. Pold' Pnew: . however.total Pold Told new old Qcell cv Mf.122 gradients. c : Specific heat of the fluid at constant volume. The density change of liquid. depends primarily on the temperature. new Pnew RT new (5. not on the pressure because of a high modulus of elasticity.1) T Pnew = old + Ttotal c v M f. The equation of state of an ideal gas is given in the following.cell where Pold' Pnew: Density of the fluid at old and new time steps. temperature and pressure are equally important parameters in determining the flow field.
5. .3 Heat Transfer Modelling The heat transfer to the obstacle is modelled as a natural convection from a vertical surface. k = c /C for liquid and 1. Qcell total : Heat input to the fluid in one cell and in the total domain.2 Energy Convection Another difference between liquid and gas is the ratio of the Specific heats.123 Mf. 5.1. The energy convection will be underestimated by a factor of 1.1.0 Monatomic. The Nussett and Grashof numbers here are based on the axial length of one computational mesh.4 for air. 5. An experimental correlation is given in Eq.. The value of k is 1. . but they may also be based on the vertical height of the obstacle.cell. 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.2 in terms of the dimensionless numbers. the atomic gases have different values of k.4 when internal energy is used for enthalpy in air flow.total : Mass of the fluid in one cell and in the total domain. convection energy transfer should be in terms of enthalpy instead of internal energy. diatomic and polyTherefore.
h = 1.124 Nu = C(GrPr) n ( 5.2) 3 2 Gr = gS(AT)L p 2 (5.708 (air) GrPr 105 .555 0.109 > 109 C 0.021 n 0.3) Pr = 0.4 The following data are used to calculate the heat transfer coefficient at the obstacle wall.923 x 103 Therefore. This may be a typical value for the natural convection heat transfer in the containment without any phase change.2 ft/sec' 5 V = 1.876 The heat transfer coefficient is given as.4 lbm/ftsec Btu/ft 2 F.0763 ibm/ft 3 g = 32. L = 2 ft AT = 500 F p = 0.8537 x 10 .25 0.7134 x 108 Nu = 89.2179 x 10 B = 1. .sec . Gr = 9.
Heat transfer area between the obstacle and fluid over one computational mesh. Heat transfer coefficient between the fluid and obstacle.2. can be obtained from heat balance between the obstacle and fluid as follows.4) where CQ is the source term in the energy equation of fluid. T = 677. Obstacle density.4 Turbulence Modelling Laminar flow is assumed in obtaining the results in Fig. the heat . 5. (Mc )ob (MC) obf hApo b ob f (5.1 and Fig.5) where M A : Mass of one computational mesh.4 sec. If the flow is turbulent. 5. pf Pob: c h : Fluid density. Specific heat at constant pressure.1. T.125 The VARR input [21) requires the time constant of the heat transfer instead of the heat transfer coefficient. CQ = Cpob (TT) (5. Therefore. 5. The time constant.
The skew differencing and corrective Huh's formula schemes are tested for various problems.126 transfer. steadystate problem in Fig. 5. 5. The flow in Fig.1 Truncation Error and Crossflow Diffusion The truncation error diffusion is compared with the crossflow diffusion in a two dimensional. 5.3(A) is parallel with the mesh orientation so that no crossflow diffusion occurs.3(B) is skewed at 450 to the mesh orientation and both truncation error and crossflow diffusions are included in the solution error.7 show that excessive numerical diffusion has occurred due to the .3.6 and Fig. 5. 5. is compared with De Vahl Davis and Mallinson's in predicting the magnitude of crossflow diffusion. 5.5 show that the numerical solutions is close to the analytical solution and the truncation error can be neglected. The results in Fig.2 Numerical Diffusion The truncation and crossflow diffusion errors are compared in a simple geometry where an analytical solution can be obtained. The flow in Fig. 5.2. The corrective scheme is tested in recirculating flow problems for mesh point and mesh interface implementations.4 and Fig.to the obstacle will be enhanced and the diffusion transport of momentum and energy will also be enhanced in the containment air. Figure 5. 5.
5. Problem geometry for the two cases.07 (B) =0.0 ay AB=BC=5. (A) parallel flow and (B) cross flow.127 (A) . for evaluation of numerical diffusion .0 a= 1.3.0.0 B T_ 0.0 (A) (B) AB=BC=5.0 u = 10.0 x BT  @x a( aT + 2 2 a2 ay ) Th= 10.0 B T specified Fig. 0 ST By =0.0 u = v = 7.0 BT S=0.
3(A) .4. 5. spatial coordinate Fig.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. 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.~ 129 Analytic solution Numerical solution with Donor Cell C spatial coordinate Fig.^XI ill I~LX_L~LI. 5.__~ . 5.5.3(A) .~ I~.__LL l.
5. analytic solution with increased diffusion constant and numerical solution with donor cell differencing convection term in 10x10 meshes along the of line CA in Fig. Comparison of the analytic solution. 5.6.3(B) .130 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Df cf T 10 spatial A coordinate Fig.
131 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Dcf T 10 / I / / ./ Fig. .J 0 I' C spatial A coordinate Fig. 5. 5. 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. 1/ 5.7.3(B) s' / line CA in .3(B) .. Comparison cT the analytic solution.
Figure 5. solution. 5.8 has the arbitrary angles of 6 and 61 such that 6 = 600 and 61 76. De Vahl Davis and Mallinson's and Huh's formulas give the same crossflow diffusion constants for the case.6 and Fig.3 Skew Differencing Scheme Two skew differencing schemes.11 and Fig. It can be seen that the former result is much better than the latter.12 for a simple pure . 5. more diffusion has occurred in Fig. 5. Figure 5. 5.7 show that the numerical solution can be reproduced by the analytical solution with increased diffusion constants by the amount given by Huh's formula.9 is the result by Huh's formula and Fig.54.6 are for 5. 5. 5. 4. meshes. 5. Since the corrective scheme is based on the validity of those correction formulas. Figure 5.53 and Eq. 4. are tested in Fig.7 than in Fig.2.2. it is necessary to test them for various cases. 5.2 Validation of Huh's Correction Formula The crossflow diffusion can be predicted by Huh's formula. Raithby's and Huh's.6. 5. 6 = 6 = 450. The problem geometry in Fig. 5. Eq.10 by De Vahl Davis and Mallinson's for the problem in Fig. 5. 5. 10 x 10 meshes and Fig.7 are for 6 x 6 It is seen that a finer mesh gives a better Since the crossflow diffusion constant is pro portional to the mesh spacing.5 and Fig.4 and Fig.8.810.132 crossflow diffusion.
133 Yo 2 T=0 D a = 1.8. Problem geomctry with arbitrary angles of 6 and 81 to compare De Vahl Davis and Mallinson's and Huh's formulas for prediction of crossflow diffusion constant .6603 DDM= 0. 5.810 8 = 600 8 Meshes 24 Meshes Fig.335 Huh A e1 = 76.0267 u = 5.0 Ax = 0.7674 DH u h = 3.0 v = 8.2406 by = 1.
134 Analytic solution with D+Dcf Numerical solution with Donor Cell T 10. 5. 5 0 spatial coordinate Fig. Comparision of the numerical solution with donor convection term and cell differencing of analytic solution with increased diffusion constant. (D+Dcf). by Huh's formula along the line BD in Fig.8 .9. 5.
8 . solution (D+DDM). Comparision of the numerical solution donor cell convection term and analytic differencing of with increased diffusion constant. 5. 5. by De Vahl Davis and Mallinson's formula along the line BD in Fig.135 Analytic solution with D+DDM Numerical solution with Donor Cell T 101 spatial coordinate Fig.10.
63 87.75 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 34.38 50 31.75 25 U 12.25 0 Skew differencing by Huh (True solution) 100 100 100 100 50 0 0 0.11.75 81.25 65. 5.5 68. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with e = 45* . 0 100 100 100 100 100 50 100 50 0 50 0 0 V Fig.5 0 6. Comparison of the true solution and solutions by donor cell scheme.25 18.136 Donor cell 100 100 100 100 100 93.
25 53.5 0 0 63.25 1.43 7.74 38.56 1.52 1.430 0 50 0 Comparison of the true solution and solutions by donor cell scheme.25 37.3' 40. 5.33 70.5E 25.8922.6 0 Skew differencing by Huh 100 90. 62.70 4.1 6.11 3.88 81.99 10.3: 11.59.01 55.2! 90.85 0.430 .12.5C 100 100 50 Fig.26 80.86 16.91 31.A Skew differencing by Raithby 100 100 100 100 96.11 0 11.22 9.25 0 0 0 .50 2.83 72.71 20.96 17.13 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 6.67 5.3E 21.9 33.137 Donor cell 100 100 100 100 50 .E 21 0 12 True solution 3.47 23. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure covection problem with e = 63.6.91 71.
to obtain the true solution by subtracting the additional crossflow diffusion . The details of those schemes are Figure 5. 5. The purpose of the corrective scheme is.2.11 has a flow given in sections 4. 5.4 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.14 show that the analytical solution can be reproduced by Huh's skew differencing scheme for the case.3.3. Fig. the However. but less than inthe donor cell scheme. both of them are conservative in a unidirectional flow. Raithby's scheme is an Eulerian Therefore. however.430 angle and neither Raithby's nor Huh's scheme can reproduce the true solution.12 has a flow at a 63.13 and Fig. 5. Some crossflow diffusion has occurred in Huh's scheme. It is indicated that Raithby's scheme may give unphysical results in problems with steep gradients of the quantity under consideration.1 and 4.138 convection problem. 8 = 1e= 45*.2. 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. method while Huh's scheme is Lagrangian. former is conservative while the latter is not. Figure 5.
13. Comparison c f the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 10x10 meshes along the line CA in Fig.139 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh C A spatial coordinate Fig.3(B) . 5. 5.
5.14.140 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh 10 5 0 spatial A coordinate Fig. 5.3(B) . 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.
In Fig.141 constant from the total diffusion constant of the finite difference equation. The full correction of the crossflow diffusion gives the true solution without any numerical diffusion error.18 show that the crossflow diffusion can be eliminated by the corrective scheme and that a finer mesh spacing always gives a better solution. 5. The flow is at a 45* angle with respect to the mesh orientation.21 gives the true solution. Two implementation strategies of the corrective scheme. Both of them are tested in a recirculating flow problem with pure convection. Fig. mesh point and mesh interface.16 has a flow with 8 = 63. 5. The corrective scheme is also tested in the problem geometry of Fig. Figure 5. The corrective scheme gives a physically reasonable solution.430. although not identical to the true solution. 5.3(B). Figure.20 shows the diffusion corrections of the mesh point and mesh interface implementations at each interface. 5. 5. The mesh point implementation gives an unphysical solution . 5. donor cell solution and the solutions by mesh point aand mesh interface implementations. Fig. Figure 5.15 numerical solutions are given for a simple pure convection problem with various diffusion corrections.17 and Fig. are introduced in Chapter 4.19 is a hypothetical 3 x 3 recirculating flow field where the inlet boundary values are specified on the left and bottom surfaces.
Donor cell solutions with various diffusion corrections for a pure convection problem with e = 450 .0 2.08 5.0 8.99 D=0.0 2.88 5 4. 0 V ________ D=0.92 5.29 5.0 10 7.142 D=0.59 7.0 2.01 7.41 u=v=l Ax=Ay=l D =D =0.25 0 5.71 5.1 10 8.5 (True Solution) 9. 5.75 6.12 5.15.88 10 5 0 2.5  3.13 7.3 D=0.29 0.71 0.5 x y Fig.0 1.5 0 1.
07 10 3.430 u=l.07 0.91 0.11 0.11 True Solution V 63.11 6.16.04 4.10 9.56  2.82 0. 5.41 0.07 2.96 3. v=2 x= y=l 2 D =D = x y 3 Fig.93 10 5.33 1.143 D=0. True solution and donor cell solutions with and without diffusion correction for a pure convection problem with e = 63.73 10 7.430 .37 3. 0 D2 3 0.59 1.
5. Comparison of the analytic solution.144 Analytic solution Numerical solution with donor cell Solution by corrective scheme 10 A spatial coordinate Fig.17. 5. numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig.3(B) .
3(B) . 5.145 Analytic solution Numerical solution with donor Solution by corrective scheme 10 5 0 spatial coordinate Fig. numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig. 5. Comparisoin of the analytic solution.18.
5. 3 ) 2 3 2 Lx = Ay = 1 Fig. Simple recirculating flow field for test of the implementation strategies of the corrective scheme .19.146 0 Ofr 3 ~LC~' / 10 1 1 5 0 1 3 3.
75 1.75 0 0 0.19 .75 1.25 mesh point implementation 0.20. 5.2 0.625 0 . 5.5 0.75 0.147 0.2 0 0 0 mesh interface implementation Fig.25 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.125 0.625 0.5 0 0 5 0 0 .75 1.
21 7.86 12.51 10 5 0 Fig.94 9.10 9. Comparison of the true solution.16 6.18 7.82 8.21 mesh interface implementation 0.08 9. 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.148 True solution Donor cell mesh point implementation 5.21.26 8. 5. 5.83 6.19 .16 11.10 8.75 6.
0. 5.3 sec with the corrections in Fig.27 is physically reasonable and even better . respectively. 5.2 kg/sec.19. 5. 5. Figure 5. 5. 5.24 in which the source mesh has the highest steam concentration. There was initially air in the containment and steam is introduced in the source mesh at the rate of 0.23.25 and Fig.26 show the diffusion constant corrections of the mesh point and mesh interface implementations and Fig.25 and Fig. 5. 5. Since the steam concentration distribution and flow field are given at the start of the calculation. After a while a flow field is set up in the This containment by the input mass and momentum of steam. The steam concentration distribution at the beginning is given in Fig. Another test calculation is done for a two dimensional. the steam concentration for the new time step can be calculated in an explicit way.26. 5.149 while the mesh interface implementation gives a physically reasonable solution.28 is not a reasonable solution while the mesh interface implementation result in Fig.22 shows the flow field in the containment and more detailed information about the flow field is given in Fig. 5. flow field is used here to test the mesh point and mesh interface implementations of the corrective scheme.28 show the steam concentration distributions after one time step. rectangular containment with an arbitrary recirculating flow field. Figure 5.27 and Fig. The true solution cannot be reproduced because some of the information about the flow field has already been lost in Fig. The result in Fig.
5.5m Ay = 1.0m Fig.150 Ax = 0.22. 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 .
2435 0.1139 0 . 2214 velocity [m/sec] Ax = 0.0127 0.2216 0. Detailed flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .93 2.7126 0. 3563 0.0937 0 1.2190 0. 5955 SOURCE 0.3 095 1.0492 0 .34 0. 861 1. 0735 0. 5.3957 0.0348 0. 0648 1.2629 0.3253 0.207 2.456 0. 045 0.3074 0.1229 0.1385 0.4870 0. 1317 0.151 0.531 0.1422 2148 0.1276 0 .5174 1.5 m Ay = 1 m Fig.2460 0.2474 04 1379 0.4920 0 .5893 0.23.2159 0.
1244 0. 5.24.0031 Ax = 0.1712 0.2 kg/sec of steam Fig.0346 0.1859 SOURCE 0.1592 0. Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .0011 0.1420 0.0049 0.0819 0.5m by = Im Source: steam concentration in Ikg/m 3 0.1710 0.0179 0.0002 0.1085 0.0136 0.0305 0.1750 0.0007 0.0095 0.1266 0.152 0.1178 0.0298 0.0620 0.0358 0.0676 0.
stant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .1452 0.0590 0. 0152 0. 0. 0227 0.1451 0.153 0.1 389 0. 0..0224 0.1279 0. 0086 0.0 362 0.0 183 Ax = 0. 0396 0.0348 0.1701 0.3262 0.0300 0. Diffusion co.0731 116 0. 0488 0.1583 0.5m by = Im Diffusion constant in Im2/sec] Fig.0 278 0.1178 0.2740 0.0794 0. 5.0600 0.0193 0.0440 0.C209 0.0937 0.1014 258 0.0311 0.0311 0.1694 3106 0.0253 0.0255 SOURCE 0. 0.25. 2422 0.1763 0.1854 0.
0192 0 0.9769 0 0 0.0479 0. 0706 0.2317 0 0. Diffusion constant corrections of the mesh interface implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .0632 0.1026 0 SOURCE 0 0 0.5m by = lm Diffusion constant in Im2/sec] Fig.0518 0 0 0.4732 0. 5.0 111 0 Lx = 0.0 923 0 0.0443 0.0 257 0.1149 0 0 0. 0240 0.2529 0.26.3694 0.0215 0.0534 0.2134 0.0287 0.154 0 0 0.0 76 0. 0579 0.0860 0.
1243 0. 5.0107 0.155 0.0003 0.0046 Lx = 0. with donor cell differencing convection term without any diffusion of correction .0016 0.1321 0.0233 0.1706 0.1363 0.0448 0.1711 0.0342 0. 0.0947 0.1847 SOURCE 0.0342 0.0068 0.0195 0.1733 0.0833 0.3 sec.1475 0.1623 0.0465 0.2 kg/sec of steam Fig.0011 0. Steam concentration distribution after one time step.5m Ly = Im steam concentration in [kg/m 3 ] Source: o.0735 0.1135 0.27.
1747 0.0029 0.0068 0.0010 0.5m Ly = Im steam concentration in Ikg/m 0.0937 0. Steam concentration distribution after one time step.0709 0.0465 0.0815 0.0092 0.1334 0.1911 0.0229 0. with mesh point implementation of the corrective scheme .1247 0.2 kg/sec of steam Source: 3 Fig.2086 SOURCE 0.0433 0. 5.0250 0.0355 0.1797 0.28.1493 0.0287 0.1646 0.0042 Lx = 0.156 0.1240 0.1729 0.0005 0. 0.3 sec.
0815 0.0424 0.157 0.1342 0.2 kg/sec of steam Fig.0228 0.3 sec.0058 0. 5.1487 0.1847 SOURCE 0.0360 0. 0.0010 0.1310 0.1756 0.0007 0.1738 0.1794 0.1243 0.0100 0.0039 ax = 0.0295 0.0947 0.5m by = Im steam concentration in [kg/m 3 Source: 0. with mesh interface implementation of the corrective scheme .0481 0.29.0192 0.0003 0. Steam concercration distribution after one time step.0722 0.1628 0.1112 0.
30. The results of the ADI and explicit schemes are compared in the problem geometry of Fig.5 sec. Figure 5. In Fig.30 there is an inflow from the bottom surface at the velocity of v=1. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable. It is found that a wise use of the ADI scheme can reduce the overall computation time in comparison with the explicit and implicit schemes.0 and the inlet boundary value of 4 is. Figure 5.0.32 shows the profiles of 4 at the same loca tion for the time step size of 1 sec. 5. it has the maximum time step size that can give a physically reasonable solution. when the Courant limit is 1 sec.31 shows the profiles of 4 along the dotted line for the time step size of 0. =10. 5. 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.3 ADI Solution The ADI scheme is tested to use a time step size larger than the Courant limit. The ADI and explicit schemes give almost identical results during all the transient. which is exactly the .158 than the donor cell solution. In the meshes at the top of the source mesh. 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.
Problem geometry for comparison of the explicit and ADI solution schemes .0 c= 1.0 Ax = Ly = 1. 5. v = 1.1I I _~ 1_.~ ~ILI_~_~LI*L~l*IIIU*XI 1 9 IB Source of ' 'A Profiles of 4 = 10.YXI~^.0 on this line are compared.0 Fig.30.
0 12.160 12.0 10.0 sec for Lt=0.0 Fig.0 11.30 the profile of sec and the Courant limit of 1.0 11.0A 12.0 11.01 12. Comparison of the explicit and ADI solutions for along the line AB in Fig.0 11.31.0 12.0 0. 5.0 10.0 10.0 11. 0A Explicit ADI 10.0 A iatlai B 10.5 sec 12.0 coordinate 11. 5.0 10.5 .
0 10.0 10.0 16 sec 20 sec 12.0 ~ spatial coordinate 12.0 sec .0 12. 5.161 1 sec 12. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig.0 11.0 10.0 sec and the Courant limit of 1. 12.0 10.0 .0 Explicit ADI 10.0 12.0 11.0 Fig. 11.0 4 sec 11.30 for At=1. 5.0 11.32.
33. Initially. No crossflow diffusion corrections are involved in the results of Fig.31. there is some deviation between the two results. 5. Fig. 5.162 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. 5.32 and Fig.33 shows that the ADI scheme can use a time step size up to five times the Courant limit. Fig. 5. . but it damps out rapidly and the ADI scheme gives almost identical results with the explicit scheme.
30 with different time step sizes .0 t spatial coordinate B Fig. 5.5 sec ec 11. Comparison of the ADI solutions for the profile of at t=10.33.0 At =0.0 sec along the line AB in Fig.163 12. 5.
energy. truncation error and crossflow diffusion. The total diffusion con stant is the sum of the laminar and turbulent diffusion constants. Crossflow diffusion occurs due to the donor cell treatment of the convection term in a multidimensional problem. in the finite difference donor cell treatment of convection. The con tinuity/momentum equations are decoupled from the scalar transport equations and solved by the SMAC scheme to obtain the flow field. The mass diffusion. 6.164 CHAPTER 6 CONCLUSION 6. or superheated steam as may be appropriate.2 Numerical Schemes (1) There are two numerical diffusion sources. (2) The models of laminar and turbulent diffusions in Chapter 3 may be adequate for predicting the hydrogen transport in the containment. The thermal equilibrium is assumed and the phase change occurs to maintain 100% relative humidity. The . and turbulence equations are solved separately using that flow field.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.
the corrective scheme is generally preferred to the skew differencing scheme. most of the numerical diffusion error in a multidimensional. recirculating flow problem is due to the crossflow diffusion. have been tried in order to eliminate the numerical diffusion. (2) Two types of approaches. but not always. Therefore. The corrective scheme is based on the fact that the additional crossflow diffusion can be predicted theoretically for every mesh at every time step. The 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. The inclusion of corner points complicates the matrix structure and a fully implicit scheme should be used for maintenance of stability.165 effective diffusion constants of the truncation error and crossflow diffusion are of the same order of magnitude. It gives unphysical results for most recirculating flow and coarse mesh problems. It is conservative and an explicit scheme can be used without affecting the simple solution structure. ~unx. . skew differencing and corrective schemes. convection dominant. 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 mesh interface implementation is always preferred to the mesh point implementation in a recirculating flow problem. The results of numerical experiments are found to be consistent with the Von Neumann analysis.2 imposes a maximum time step size that can give a physically . the physical constraint in section 4.4. are tried for the corrective scheme. are identical in a unidirectional flow. The graphical method is used to obtain the stability condition in terms of the Courant. First. dx and Rx .e. 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. mesh point and They mesh interface.166 (3) Two implementation strategies. using the characteristics of a quadratic equation. i. it has its own limitations due to the following factors.. Cx. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable. (5) The maximum time step size is limited by the Courant The Alternate Direction condition in an explicit scheme. Implicit (ADI) scheme can be used to increase the time step size and decrease the overall computational effort. (4) The Von Neumann stability analysis is applied to various finite difference forms of a general conservation equation. diffusion and cell Reynolds numbers.
asymmetry of the given problem may give different results according to the sweeping sequence. For example. the ADI scheme is a fractional step method. Second.167 reasonable solution. a steadyThe state solution may never be reached due to the fact that one time step is composed of a few fractional steps. .
velocity gradient. time step size. etc. transitional and laminar flow regimes. The flow regime criterion may be replaced with a model that covers all the turbulent. distance from the wall). to eliminate the crossflow diffusion. The model should show proper limiting behaviors as the Reynolds number goes to infinity (turbulent) and zero (laminar). . 7.168 CHAPTER 7 RECOMMENDATIONS FOR FUTURE WORK 7. The important parameters may be the velocity. A criterion for the flow regime should be developed in such a form that it may be implemented into a computer code. it is necessary to predict the flow regime in order to use appropriate physical models. geometry (e. previous time step information.1 Physical Models Since both the laminar and turbulent flow regimes are possible in a complicated geometry and flow field. multidimensional recirculating flow problem. They may still be improved further to get an accurate solution in a transient.2 Numerical Schemes There have been two approaches. skewdifferencing and corrective schemes. The modified kE model in a low Reynolds number flow may be a guide in this approach..g.
It may be possible to develop a reasonably simple form that may or may not include the corner points. The conservative form of the skew differencing scheme is complicated and time consuming in comparison with the corrective scheme. ADI and implicit schemes although only the implicit scheme has been used up to now. The stability of the skewdifferencing scheme is open to question in the explicit. 2. The solution in Appendix A may be a useful guide in this work. More tests and validation calculations are required in recirculating flow problems. Skew differencing scheme 1. .169 Corrective scheme The corrective scheme has been successful by the mesh interface implementation strategy in the scope of this work.
F. [2] [3) [4] [5) [6] [7] [8) [9) [10) [11) [12] . Boyle.170 REFERENCES [1] Amsden.. et al. Briggs.B. Phys.H. A. Baliga. Comp. B. Briley.User's Manual. P." Journal of Comp. Broadus.R. and Eng. 233241 (1975). Cebeci. and Lightfoot. 34.H." HEDLSA2682 (October 1982). and McDonald. "BEACON/MOD 3: A Computer Program for Thermalhydraulic Analysis of Nuclear Reactor Containments .G. 5473 (1980). et al. "Hydrogen Distribution in a Containment With a HighVelocity HydrogenSteam Source. G. "The SMAC method: A numerical technique for calculating incompressible flows. Bradshaw. Amsden. R. "A New Finite Element Formulation for ConvectionDiffusion Problems. and Harlow. thesis (1979)." Journal of Comp. W. S. A.. 393409 (1980). "A Finite Difference Scheme for the Incompressible AdvectionDiffusion Equation. Phys. W. 6. "On the Structure and Use of Linearized Block Implicit Schemes." NUREG/CR1148 (EGG208). and Harlow." LA4370 (May 1970).R.. Stewart.V.A.. Briley.R. Bloom.D. A." Numerical Heat Transfer 3. H.E... H..H.. 3...R. Phys. and McDonald.. Academic Press (1981). 94110 Amsden. and Whitelaw.. E. D. and Patankar. Bird. "Transport of Turbulence in Numerical Fluid Dynamics. 24. J. Phys.R." Journal of Comp. Transport Phenomena. C. and Harlow.A.. W. "Solution of the Multidimensional Compressible NavierStokes Equations by a Generalized Implicit Method." Comp. 372397 (1977). 332325 (1970)." MIT Ph. "Transient Effects in Turbulence Modelling. Methods in Applied Mech." Journal of (1968).R. "A Simplified MAC Technique for Incompressible Fluid Flow Calculations. T. 6.. F.H. John Wiley & Sons (1960). F. D. (April 1980).N. Engineering Calculation Methods for Turbulent Flow.A.
Method for Convection in Numerical Fluid Dynamics. thesis (1977).J. "Simulation of Containment Cooling with Outside Spray After a Core Meltdown. Keller. 1976). Clark. Varadi.C. and Ramshaw. of Fluids Supplement 2 (1969). "SecondOrder Numerical Advection." CRBRPARD0106. Chang. Chang. Phys.I.. J.K." The Phys.S. 7179 (1979).. C.." MIT Ph. "Study of Nonhomogeneous Conditions in Reactor Containments Due to Separation Effects. 27. thesis (1981)..F. G." MIT M.. J. "Comparative Analysis of Numerical Methods for the Solution of NavierSTokes Equations. M. S. "Buoyancy Effects in Mathematical and Numerical Turbulence Modelling.P. "Tensor Viscosity Dukowicz.B. 2943 (1976). and Mallison.S. [14] [15) [16] [17] 118] [19] [20) [21] [22] 123] [24] [25) . and Design 69. Vol. and Eng.W. "SOLAICE: A Numerical Solution Algorithm for Transient Compressible Fluid Flows.. Jr..H. K. De Vahl Davis. S. R. B." LA6236 (July 1976).. "Numerical Methods of Reactor Analysis. Williams.. N. Hirt. Cook. 127137 (1982). T. L.D.. P..171 [13) Carbajo. 1344 (1981).G. and Nakayama. Hirsh. and Hansen. Covelli. Academic Press (1964). and Design 69." Journal of Comp." MIT Ph. Crowley.. W.. "Coolant Mixing in the LMFBR Outlet Plenum. 1 (Nov. J.L.. Cebeci. Nielsen.I. M." Journal of Comp.D. H. Phys.B. Vol. J. 32.D. P. 95107 (1982). Cheng. Methods in Applied Mech..A Computer Program for Calculating Timedependent Turbulent Fluid Flows with Slight Density Variation. Cloutman. S. "Accuracy of Difference Formulation of NavierStokes Equations. "VARR2 ." Nuclear Eng. "Studies of Numerical Methods for the Plane NavierStokes Equations. thesis (1980). G..H. 1.. Chen. 4." Comp..D." Nuclear Eng. and Lewis.. "An Evaluation of Upwind and Central Difference Approximations by a Study of Recircul ing Flow." Computers and Fluids. L..D. Y. 471484 (1967). and Romero. G.
.W. Phys. [27] [28] [29) [30) [31] [32) [33) [34] [35] [36) [37) Hirt.The 1978 Freeman Scholar Lecture. Hirt. J.. Molecular Theory of Gases and Liquids.C.. 3.. Harlow...H.A. 29.W. J.W.. 8. New York (1954). Curtiss. Phys. 101. Phys. "Buoyancy Induced Fluid Motions Characteristic of Applications in Technology ." Journal of Fluids Eng. and Marohar. T....H. "Numerical Calculation of Multiphase Fluid Flow. 557573 (1980). Gupta.. and Amsden. Harlow. G.E. Turbulence.P.A. M. F. 17.. F. B. and Amsden." The Phys. C. Vol. Humphrey. 1952 (1975). "Practical Investigation of Convective Difference Approximations of Reduced Dispersion. R.A. 8093 (1968). of Fluids. McGrawHill Book Company. 309312 (March 1975)." The Phys." Comp.F. 18. [38) . "A Numerical Fluid Dynamics Calculation Method for All Flow Speeds. Methods in Eng. "Improved Form of the Low Reynolds Number kc Turbulence Model. F. 339355 (1968). C. Methods in Applied Mech." Journal of Comp.. M. "A Comparison of Hybrid and Quadratic Upstream Differencing in High Reynolds Number Elliptic Flows. "Computer Studies of Timedependent Turbulent Flows. Phys. 2. Hoffman. and Eng.O.. C. and Launder." The Phys. Hirschfelder. for Num. 2nd ed. of Fluids Supplement 2 (1969). A.. Gebhart..E." Int.M.H. A.A.." Journal of Comp.H. R. R.. (1975). and Amsden. "Heuristic Stability Theory for Finite Difference Equations. 15. and Bird. A. B. 528 (March 1979)... Wiley. 197213 (1971). Han." MITEL 81020 (June 1981).B. of Fluids Supplement 2 (1969). Vol. J. "Behavior of Buoyant Moist Plumes in Turbulent Atmospheres. J. Vol. "Numerical Calculation of Almost Incompressible Flow. J." Journal of Comp.O." Journal of Comp. Hinze. Harlow.. Hamza.. 8195 (1981).172 [26] Fromm. "On the Use of Central Difference Scheme for NavierStokes Equations. and Golay.
" J. Methods in Applied Mech.K. Battelle Institute e.G. Launder. Fluid Mech. Lam. 301314 (1972)." Trans. W. translated (October 1980).B. F. Manno." Int. 221239 (1974).. 68. and Spalding." BF63. 103. M. 456460 (Sept. and Launder. 11191130 (1973). V.173 [39) Houston. and Rodi. "The Calculation of Low Reynolds Number Phenomena with a Twoequation Model of Turbulence. John Wiley & Sons (1978).E.E. W.. Methods in Applied Mech. "Progress in the Development of a Reynolds Stress Turbulence Closure. G. and De Braemaecker.D. [40) [41) [42) [43] [44) [45] [46] [47] Leschziner. "The Numerical Comparison of Turbulent Flows. Keenan. Manno. Jones. 1981). Launder. of the ASME... and Eng.. C.. et al.3633. Jones.J. "Practical Evaluation of Three Finite Difference Schemes for the Computation of SteadyState Recirculating Flows. W. Heat Mass Transfer. J. 16.P. Langer. and Launder. part 3.. Phys." Journal of Comp. and Bremhorst. B.. B.A.. Vol." MIT Sc. K. and Golay. 15. 269289 (1974)... G.. M. thesis (1983). Vol. J. B.E. and Keyes..G." Int.H. 293312 (1980). "Experimental Investigation of Hydrogen Distribution in a Containment of a Light Water Reactor Following a Coolant Loss Accident. 537566 (1975). 16. Vol... J..P. 23.P. V.W. "The Prediction of Laminarization with a Twoequation Model of Turbulence. Frankfurt. M. "Analytical Modelling of Hydrogen Transport in Reactor Containments. Cl." Comp.." Comp. Reece.E. J.V. B. "A Modified Form of the kc Model for Predicting Wall Turbulence.P. Jr. 3. "Analytical Modelling of Hydrogen Transport In Reactor Containments.H. and Eng. "ADI Solution of Free Convection in a Variable Viscosity Fluid. D. Heat Mass Transfer. 148) [49) . Vol." Report submitted to 2nd International Topical Meeting on Nuclear Reactor Thermalhydraulics (March 1982). Steam Tables.
4. Methods in Eng. Inc." Int.. S. J. 17871806 (1972). 541550 (1972). Mass and Momentum Transfer in Three Dimensional Parabolic Flows... 10. W. and'Eng. K. "A Note on the Empirical Constant in the KolmogorovPrandtl EddyViscosity Expression.." Mathematics of Computation. Rodi. Turbulent Buoyant Jets and Plumes." Journal of Comp. for Num.. 153164 (1976). Computational Fluid Dynamics. Raithby. McGrawHill. "A Calculational Procedure for Heat. Roberts. N. K..J.. Reid. Phys. Patankar. of Fluids Supplement 2 (1969). Hermosa Publishers. Pergamon Press (1982). Hemisphere Publishing Corporation (1980). "Upstreamweighted Differencing Schemes and Their Application to Elliptic Problems Involving Fluid Flow. 191206 (1974). New Mexico (1976)." Comp. 169184 (1972). A.. P.O.. Methods in Applied Mech. 2. P. of the ASME. R. 1975).B. 9. 15. S.V. John Wiley' & Sons. Raithby. 9. and Sherwood. "Numerical Experiments with a Difference Model for the NavierStokes Equations (Turbulence Model).K. "On Artificial Viscosity.C..." The Phys. Roache. [52) [53) [54) [55) [56) [57) [58) [59] [60) [61) [62] [63) ." Int.. and Torrance. D. Heat Mass Transfer. Schonauer. Methods in Applied Mech." Computers and Fluids.B. Patankar. W. and Eng. Runchal. S... Computational Methods in Engineering and Science.. G. 75103 (1976). J. (1977). 20 (1966). G. Rodi. "Convergence and Accuracy of Three Finite Difference Schemes for a Two Dimensional Conduction and Convection Problem. Roache." Comp. Raithby. "A Critical Evaluation of Upstream Differencing Applied to Problems Involving Fluid Flow. New York (1958). T..K.. W. "SkewUpstream Differencing Schemes for Problems Involving Fluid Flow. and Spalding.." Trans. Vol.174 [50] [51) Nakamura.V.D.D. The Properties of Gases and Liquids.E. G. Albuquerque. Numerical Heat Transfer and Fluid Flow.J. and Weiss.D. "Convective Difference Schemes. 386389 (Sept.
95102 (1950). G. Stuhmiller. Phys.. 1981).. 72B.B. G.." of the National Bureau of Science . and Eyler. C.. 4. Methods in Applied Mech. Wilke. J. K. "Comparison of Finite Difference Journal of Computations of Natural Convection." Comp. Y. J. L.B... Stubley. 20. D. "TEMPEST: A Threedimensional Timedependent Computer Program [72] for Hydrothermal Analysis..J. 2. 3. . 3. A. Tennekes. "Simulation of Convection and Diffusion Processes by Standard Finite Difference Schemes and by Influence Schemes. 1974). 35.B.W. 220237 (1976). H.L. Smith. "Proposal for a New Discrete Method Based on an Assessment of Discretization Errors. 551559 (1972). MatheResearch matical Sciences.D. G." Chemical Engineering Progress. No. Strong. and Valentin.D." Journal of Comp. 46.E. and Hutton." SAI74509LJ. "Numerical Analysis of Fluid Hammer Waves by the Method of Characteristics. K. Raithby. A. "Development and Validation of a TwoVariable Turbulence Model. (Jan. 411428 (1980). "Diffusional Properties of Multicomponent Gases." FATE80114 [73] (Jan." Numerical Heat Transfer. Budden. D." Int. R. 411428 (1980). A.A. 1968). M.175 [64] Shin. Raithby.G. Stubley. J. A First Course in Tur [65] [66) [67] [68) [69) [70] [71] Torrance... Trent. and Lumley. G.. bulence..Dec.. Spalding.. Methods in Eng..B.M. 153168 (1982). 4 (Oct. No. and Eng. R.R.. "The Numerical Treatment of Advection: A Performance Comparison of Current Methods. Vol. Vol. "A Novel Finite Difference Formulation for Differential Expressions Involving Both First and Second Derivatives.S.A. MIT Press (1970).H. and Strong..D. For Num.L." Numerical Heat Transfer.D.. and Woolner.
2 x 2 a2 ay = c 2 (A.1 is2 reduced to the following form. as follows.a2T aT 2 + a2 2 2c ax ax 2 y ay (A.1.3) ax .2) Equation A.3(B) A steady state two dimensional energy conservation equation without any source is given in the follbwing Eq. A. c. Eq.1 convection occurs in the x direction and diffusion occurs in both x and y directions. a= k PCp C =2 u 2A. A.3(B) and the constants can be grouped into one constant.2 is simplified further by substitution of 4 for T as follows..1) In Eq. The boundary conditions are specified in Fig.176 APPENDIX A ANALYTICAL SOLUTION FOR THE PROBLEM IN FIGURE 5. pc u aT = k(a2T + 2aT 2 x2 p x ax ay (A. 5. Therefore.
3 is reduced to the following form. A.8) dy where a > c > 0 .y) = X(x)Y(y) (A.6 should be equal to some constants because they are functions of x and y only respectively and the sum of them is equal to a constant. I d2X + 1 d 2Y 1 = c2 C (A.177 where 4 = eCXT (A. which are functions of x and y only respec (x. The variable 1 is separated into two vari X and Y.a =  (A.6) dx dy The two terms on the left hand side of Eq. A.4) The boundary conditions should also be expressed in terms of 4 as given in Fig.5) Then Eq.7) d Y c . A. tively. Equation A. c 2 X cx 2 = (A.1. ables.3 can be solved by the method of separation of variables.
5.178 .1.3(B) where 0 is given by.= YO T 0 ii ii i h YO + co = 0 4 = 0.0 0 = 0 x Fig. Boundary conditions in terms of 4 for the problem in Fig. A. 4 = ec T .
A. dY dy y=O Y=O 0 (A. the constants in Eq.179 The solutions to Eq. A. C3 = 0 (A.12) (d dx _cX) = 0 x=xO (A.9 and Eq.7 and Eq.10) The boundary conditions in Fig. Eq.1 may be given in terms of the variables X and Y as follows. A.16) . 13.10 should satisfy the following relations.14) (A. X = Cleax + C2 eax (A.9) Y = C3sin8y + C4cosSy (A.8 can be easily obtained as follows. 13) In order to satisfy the boundary conditions given in Eq.15) ny 0 = n C2 _ n+c 2 e nX 1 n (a n 9 C) (A. A. A. The boundary con dition at x=0 will be considered later.11.11) dY dy=Y 0 = 0 (A. A.12 and Eq. A.
) (A. 17) T(x.19) . .2.18) n n Since the boundary condition at x=0 should also be satisfied. 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. Therefore.0 2 A n Th h nr 1 n 1+( a c )e n 1 a +c 2a x . the final solution can be expressed as follows.180 where 2 =c 2 + n (O) n 2 yO =c 2 + 8 2 . nw sin2 nO0 (n=1. the constants A.1 are determined as follows.y) where nr (A. . Th 0 2.y) = e cx (x.3 .
The sequence of the sweeps is arbitrary and depends on the problem under consideration. They are formulated for a natural convection problem where the gravity acts in the zdirection.3 are the z and x direction sweeps of the x direction momentum . When the tilde phase of the SMAC scheme and other scalar transport equations are solved by the ADI scheme. Since there are two steps involved in solving the momentum equation in each direction.1 and B. 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. The ADI scheme is applied to a two dimeraional momentum equation in the following Eq.1Eq. Therefore.i. Equations B. B.iillaY~i~~l_ ^YII^^~II ( . The most important sweep should be given the first priority and most updated information. B. there are six sweeps in a three dimensional problem and four sweeps in a two dimensional problem. the computational effort for one time step is greater than that of the explicit scheme.4. The time step size limitation in the ADI scheme comes from the semiimplicit treatment of the nonlinear convection term. 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.
182 equation.4 are the x and z direction The z direction sweeps of the zdirection momentum equation.2ui+. Equations B. 2 Az + Ui+ 1 P  n  i+j i+ jui+4j (B.P z  n * RZij+ ij+ ij+ (B.2ii+ Ar 2 + Wilj+ . u n * u j i+ Ui+D Dt j + * n i+ j+ Ui+ j .2 and B.u n+l ui+2 2 n+l Ar2 +2u j n+l ui j . ar . sweep of the z direction momentum equation is treated as the last one in order to be given the most updated information.u Lr * n+1 .3) .1) * * n * *  ij+ At Wi+1 j+ ij+ + i+Ij+wij+ uij+Wilj+ 1 P n~ .Wi+ Az j Ui+ j 1 ui+ j+l * .ui+ j At  n+l * u i+lj i+ .Rxi+ jui+ jui+ j RX * n+l (B.2) n+l i+j * .
+ n+1 n+l Az 2wi + w _P z P0 zP g z RZ n+l ij + wij+ 3wij+ (B.4) Equations B.4 have given almost identical results with the explicit scheme for a time step size below the Courant limit. B. .183 n+l i+ * wij+ * i n+l +lWij+ * n+l zW. and B.3.2. w At n+l Wi. B.1.
5. a flow split occurs as shown in Fig. or all inflows in x. Then four crossflow diffusion constants at every interface are summed up to give a total crossflow diffusion constant . 'progl fortran'.8.4.2. The file. There are When there are inflow and outflow. Fig. 5. are the computer programs for calculating the numerical values of the analytical solution in the problem geometries. y and z directions. the calculational logic of the mesh interface implementation of the corrective scheme in section 4.3.. 4. y and z direction mesh interfaces.8. 5. 'prog2 fortran' and 'prog3 fortran'. the crossflow diffusion constants can be calculated in a simple way as in the first part of the program. When there are two outflows in any of the x.184 APPENDIX C COMPUTER PROGRAMS The files. 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.3(B) and Fig. are assumed to be zero. All the inflow velocities The component velocities are used to calculate the crossflow diffusion constants for eight subspaces in a three dimensional coordinate space. 'difprg fortran'.3(A). two different cases treated separately.
. A portion of the VARR program is introduced to show how the ADI scheme is implemented in the energy and momentum equations. the positive x direction.185 because one axis direction. The ADI scheme for the momentum equation is given in Appendix B. The subroutine 'tdm' solves tridiagonal matrix problems by forward and backward sweeps. has four neighboring subspaces. The total crossflow diffusion constant can be calculated in the same way for every mesh interface. It is used to solve a one dimensional implicit finite difference equation in the ADI scheme.
N FILE: PROG FORrRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT REAL*0 (AH.26 II IF  IP  I GO TO 55 ALN = DSORT(CSO*II*II*YSO) (II.14 55 ATOT * THIODEXP(ALN*XO)/2.0 44 YY I IIPI*(UJ0.5)/NY ADD * ATOTTDCOS(YY) SUM T(d) v SUM + * ADO * DEXP(C*XO) 20 CONTINUE SUM 10 CONTINUE WRITE(G.0*TII/PI PI/2.O) ALNX . AAN*2.3F7.NY  0.1415926536 C = U/(2.0 00 20 IPsi.04ALP) CSO .O*ALNXO)/(ALNC) COF * CCOFF/lI AAN .I=l.U.0 IHPI DO SUM 10 JJI.Tl.OZ) T(tO) DIMENSION REAO(10.COF*DSIN(II*HPI)/(1.215.OALNOEXP(ALN.3) SlOP END PRO00210 PRO00220 PROO0230 P1ZOOO240 P1I000250 PROOO260 PROOO270 PROOO200 PROOO290 P11000300 PROO0310 PRO00320 PRO00330 PROOO00340 PRO00350 PROOO360 PROOO370 PROOO300 PROOO390 PROOO400 .4ALNX) ATOT . C*C XO " NXOX YO * PROOOO0 PROOOO20 PROOO30 PROO0040 PRO000OOOO P14000070 PRO000OGO P11000090 POOOOT100 PROOOO10 PIIOO0120 PRO00130 PROOO140 PROOO150 PRO00160 PROO000170 PRO00100 P1.1 ) DX.000190 PI00O200 NY*OX YSO v PI*PI/(YO.5F11.ALP II FOIRMAT(F7.2) PI a 3.NY.100) (T(I).NY) 100 FOnMAT(IOX.EO.YO) CCOrrF 2.XO)/(ALNC) GO TO .2.IALNIC)*DEXP(2.NX.
TC(IO) TH14.0 SjU3 " = 0.(ALN#C)*DEXP(2. GH(10).5) EXAL = OEXP(ALNOX) XAL = DEXP(ALN'X) EXALN D EXP(ALN*XN) XALN = DEXP(ALNXN) IF(II.0 SUMO SUMI r 0. U.5) XALO = DEXP(ALN*XO*O.OCOS(OEN*Y4) Y4S DOSIN(3EN'Y4) ADO * AAN*(EXALO + ALNX*XALO)*YOO PROOO520 PR000530 P1OOO00540 PR000550 .U/(2.XN = XN .OTHIH/PI 11PI = PI/2.21 II = IP  I ALN = DSORT(CSO + IIII*YSQ) EXALO = DEXP(ALN*XOO.0 NY2 XXX NY/2.FO. 11 FORMAT(F7.0 SUM2 = 0.21)*DX = X 4 YO/2.X = 1.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN .NY2 0. NX.5*YO . GL(10).0 .1. IC1l REALi0 (AII.0 SUMaX SUM4 XN vI Y2 Y3 Y4 • 0.O.5)*DX = YC = (JJ. ALP READ(10.11) DX.0Z) DIMENSION TII(10).0 X = (JJO.0 X + XO/2.3F7 .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 = 0.COF*DSIN(II*HIPI)/(I.3. NY.0  RTO DO =  OSORT(XXX) 10 JJ=1.+ALNX) BEN 11*PI/YO YOO .215.p A FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 MPI. TL(10).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 1.2) PI . 1415926536 C .DCOS(BEN*YC) .2.DSIN(BEN+Y3) Y4C .DCOS(BEN*YI) YI Y22 • DCOS(BENY2) OCOS(BEN*Y3) Y3C Y3S .0 .O) GO TO 55 ALNX .0 = YO/2.0 SUM3X 0.
1OO) (TH(I).I WRITE(6.0 AOOX a 0.AAN*(EXALALNX*XAL)*Y22 ADO3X v AAN*((C+ALN).5 A002 * TMIIIXAL0.5 A)DO * TIHl'XALO*O.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 .0 ADOvY * 0.00) (GL(I).NY2) = I.0 77 SUMI = SUMI + ADO SIIMO a SUMO + A000 SUM2 a SUM2 + A002 SUM3X = SUM3X + A00O3X SUM3Y w SUM3Y + ADD3Y SUM4X s SUM4X + AOO4X SUM4Y a SUM4Y + ADO4Y 20 CONTINUE C TC(dJ) THI(JJ) TL(J) Gi(JJ) = = GL(dJ) 10 CONTINUE C WI ITE(6.tOO) (TL(l).NY2) WRITE(.II.tOO) (TC(I).NY2) = t.II.NY2) WRITE(6.3) STOP ENO SUMO*DEXP(C*XO0O.EXALN + ALNX*(CALN)*XALN)*Y3C EN*AAN#(EXALN .I 1OO FORMAT(IOX.NY2) WRITE(G6.(" % FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 ADOI w AAN*(EXAL+ALNX*XAL)*YII AD02 .5) SUMI*DEXP(C*X) SUM2*DEXP(C'X) (SUM3X+SUM3Y)*DEXP(C*XN)/RTO (SUM4X+SUM4Y).0 ADO3Y * 0.SFII.II.100) (GH(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.5 ADO3X 0.
XO/RT3.215.5)*(XOYO/1r3)/NY2 YC=(JO.0 XC=Y0/(2.0)/NY2 X2=(JJ0. U.0.5)*YO/(2. TL(IO).0*ALP) CSOQ C*C XO = DX'NX YO = DX*NY YSO = PI*PI/(YO*YO) CCOFF = 2.2.DCOS(iEN*tY2) YV3C = DCOS(BEN'Y3) Y3S = DSIN(BENtY3) YAC = OCOS(3EN*Y. ALP PROOOO10 GL( O).5) * (YO/3. TC(1tO) PROO0020 PROO0030 PROO0040 PnOOO050 PROOG000 READ( 10.C P11000220 PROO0230 PROO0240 P1000250 PROOO2GO PROO0270 P11000200 PROO290 PR000300 PROO03 10 P 11000310 PROOO320 00 PRO00330 PR00OO340 PROO0350 00 20 IP=1. I 1) DX.O*RT3)+(JJO.0(JJ0.O' XO/RT3)/NY2 XI=(dJ0. rORMAT(F7.5)YO/(2 .0 NY2 nRo P000O070 13100000 PROOOOOO PROOO0000 PRO00110 = NY/2.22 II IP I PR00360 PRO00370 PROOO3UO P(1000390 P000400 ALN = DSORT(CSO + II*laIYSO) GO 10 55 Ir(II.S)*(XO/RT3YO/6.5)*YO/(2.OtTIIIM/PI HIPI = PI/2.RT3)+(dJO.S)#(XOYO/(2.N*X2)+ALNX')OEXP(ALN*X2))*Y22 AD003XwAAN((C+ALN)*DEXP(ALN9X3)4ALNX*(CALN)*DEXP(ALNtX3))Y3C PRO000410 PROO420 PR000430 1R'000440 P1100050 PR00O4GO PROO0470 PlOOO400 POO11000490 PROOOSOO 1100005 10 PROOO520 P0000520 P1OO0530 PROOO040 P1000550 .0)/NY2 X4XOYO/(2.5)YO/(2.*0RT3#NY2) Y4=YO/3.02) 111(10). NX.DCOS(UENOYC) YIl DCOS(OEN'YI) Y22 .0+(dJO. G1(10).0*RTJ))/NY2 YIYO/2.EO. ( 10).SO5T(2.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN = COF*OSIN(II'HPI)/(I.NY2 SUMO = 0.0 PR00O170 PR000100 PROO0190 PR000200 PlO00210 SUM2 = 0.C.0 SUMI = 0.0) PROO000120 PROO000130 PROO0140 PROOO150 PR000O160 RT3 = C SORT(3.(dJ0.NX*DEXP(ALNXI))*YI I ADD2AAN*(DEXP(AI.0 SUM4Y = 0.o) ALNX = (ALNwC)tDEXP(2.0*NY2) X3YO/(2.2) PI = 3.O*NY2) .0) DO 10 00=1.3F7.0 SUM3X = 0.4) Y4S = DSIN(6ENlY4) ADDO=AAN*(DEXP(ALN*XC)4ALNX*DEXP(ALN*XC))*YOO ADDIAAN9(DEXP(ALN*XI)+AI.0 .0*RT3#NY2) Y2=YO/2.0*RT3))/NY2 Y3=(JJO.5)*(XO/RT3YO/6.0.0*RT3)+(J0.0 SUM4X = 0.1415926536 C = U/(2.+ALNX) OEN = II'PI/YO YOO . TIIl.0 SUM3Y = 0.5)*(XOYO/(2. NY. FILE: PROG3 FORIRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT DIMENSION C 11 REALtO (AH.
X4) ALNX*(CALN)*OEXP(ALN*X4))*Y4C ADO4Y.11.0 77 SUM1 .5 ADO3X PROOOSGO PROO00570 PR1000500 PROO000590 PROOOGOO P1100000.I=.OEN*AAN(DEXP(ALN*X4)4ALNX*)DEXP(ALN*X4))*Y4S GO TO 77 55 ADDOTIIII*DEXP( ALN*XC)*O.0 ADO3Y = 0.I=I.NY2) WRITE(6.I= ..0 ADO4Y = 0.100) (TC(I).I1.NY2) WRITE(6.0OEXP(CXI) PROO0670 PR000600 PROOOG90 PRO00700 PRO000710 PROO00720 PROO0730 PROO00740 PROO00750 PR 00760 PROO070 PRO00790 PRO00810 PRO00020 P'11000830 P1i000840 PROOO00OSO PRO006O PROO00070 PRO0000 PR000890 PRO00900 PROOO0910 TL(JJ)=SUM20+EXP(CX2) GL(JJ)=(SUM3X*O.5+SUM3Y*0.5 ADD2TtII#DEXP(ALN*X2)*O.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.5*RT3+SUM4YtO.100) (GL(I).NY2) WRITE(6.NY2) tOo rORMAT(IOX.SF11. 100) (GI(I).5 ADDH11H*DEXP(ALN*XI)*0.NY2) WRITE(G.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.3) STOP ENO . = 0.1OO) (TL(I). PROOO6tIO .100) (TH(I).5)*DEXP(C*X4) 10 CONTINUE WRITE(6.5*RT3)*DEXP(COX3) GHi(JJ)(SUM4X#O.0 ADDOX a 0.
DIrn*ISI .5*( WP C IF( ISA.0 RFGCR 0.VI + V2 WW = WI 4+ W2 PPX .0) IF(VP. bt ts .2.IS .IFZA(0).GT.0.LT.WM.UU/DFLR PPZ * VV/OELZ PPT .5) DELR.OR.DIFRA(0)1.DIFTA(0) REA( 10O.2) UU * UI + U2 VV .0 RFGTI .ISt + IS2 ISO IS3 + IS4 ISC ISS5 + ISG C UI = 0.VELT(O).O.) ISA .0 ) C DIFR .0 IF(UM.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 .GT.0.O.UP.VP.LT.5*( VP WI * 0.EO.ISC.0 n I FOO 130 OMTI m 0.O.I 01o00200 DIF00210 DIF00220 DIFOO230 DIFOO240 DIF00250 DIF00260 DIFO0270 DIF00200 OlFOO290 DIFOO300 DI F003 10 I'I .O.WW4DELTII#( PPX + PPZ )/PSUM C RiFGII .2.0) IF(VM.( WM W2 0.VELZ(0).DELZ.GT.5*( UP VI 0.t = 1 = i v 1 * .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 OMT .PPX $ PPZ + PPT GO TO 500 IF( PSUM.0) IF(UP.0 RFOT 2 0.0.5+( UM U2 0.WW/DELTII PSUM .O 15S3 .L1.O 151 4 0 ISS1 O 156 O DIFOOOIO DIFOOO20 DIFOOO30 OIF0040 011:O0050 DIFOOOGO DIF00070 DIFOOOTO DIF00110 RFGRI = 0.DEL'TII.0) IF(WM.O) IF(WP.OR.VM.O.5*( VM V2 0.WP 1SI * 0 152 .DIFII'IS2 rGR RfGTI .UU*DELR*( PPZ + PPT )/PSUM DIFZ * VV*DELZ*( PPX + PPT )/PSUM DIFT .O.5.UM.4 ~ FILE: O FI'IlG FOITIAN A VM/SP COINVERSATIONAL MONITOR SYSTEM PACE 001 DIMENSION VELX(0).EQ.EO.EO.
PFT u i.IPFZ)s( 1.) U2 4PF Z PET 1J2'( 1.0OPF Z) W2 *P FX' IF Z WI'.0 oirZA(l) ..0PFX)*PrET V2*( 1.NC.0PFZ)'$( . oi rooG to Dl F00G20 D1F00630 01F00640 Ut 'PFZ .0 Ir (W1#W2).0PFZ)*PFT U20( 1.Z (3 ) VELZ(4) V EL ( 5 ) 011'00720 DIF00730o DI F00740 ui roo75o DI F007GO ui FO0llO DIF 00700 0D F00190 0117OO8OO OIFOOII10 o)1roos~o OI F001130 DI FOO 40 otrooaso 01 roooGo Dl FOOD 70 VELZ( 6) VELZ(7) V EL Z( 0) VrELT( I) VELT(2) VELT( 3) VELT(4) VELT(5) VELT (6) VELT (7) VELT (0) Dl FOODO DIFOOR90 0 fF00900 01 FO009tiO 01F00920 DIFOO930 oDIF 00940 OlE 00950 Di FOO960 Dl F00970 0 F0090 01 F00990 oiro0tooo W2$( t.0.o 'FX'( 1.DELTH.V(LT(I).EQ.0.0 oiETA(I) .0PET) Vi*PFX*( 1.X(I)'DELR*( PPZ + PPT )0 'PSLJM DIFZA(1) m VELZ(I)*DELz#( rrX +.0PET) Vt .0PFX)*PFZ DO 30 11.O. 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.( 1.PFT) WI#PI'X#PFZ WI'*ix( .VELZ(I)/DELZ PPT .0.NE.( .0PFT) U2*PFZ*( 1.0PFX)o( t.0PFZ) DIFO00650 oDI roofGo oir00670 nir006s0 mrof0069 ot rooioo DIF00710o VEIX(S) VELX(6) VIELX(7) VELX(0) VEL7( I) V El.*( PPX + PPZ )/PSUM GO TO 30 600 OIFRA(I) .OPFX)O( I.0 30 CONTINUE RFGRI RFGR RFCTI  DIFO 10 tO 011:'01020 DIFO 1030 01 FOt040 DO 1050s DI FOtOGO DIFO 1070 01000 DIFO 01 FOtO0J0 OIFOi tOO DIFRA(5) v DIFfIA(t) aDIFZA(2) + DIFRA(6) + DIFRA(2) + OIFZA(3) DIFPA(7) + OIFRA(3) + DIFZA(6) + + + OIFRA(fl) OlfRA(4) + DIFZA(7) r % I .0PET) Vl*( 1.0PFX)#PFT V2( 1.0 0.o VELX( 1) VCL~X(7) VELX(3) VCLX(4 ) PFX i'rz PFT   U1/(UI'U2) Vt/(VI14V2) Wl/(W14W2) r DoIooseo 01F00590 DirooGoo.09'FX)#(t1.0. Z(2) vrI.0PFz) W2 &1EX'&( 1.0.oPF7Z)'PFT UIs .( i.0 Pr T Ir (UI+U2).0.0PFT) U I PFZ (I 1.0PFT.0 (Vl*V2).0PFZ) W20( 1.PrT )/PSUM DIFTA(I) .0PFX)'PFZ Wt'(I.VELT(I)/OELTH PSUM a PPX + PVZ + PPT IE( PSUM.0PFT) V t.VEU.VELX( I)/I)ELR PPZ .FILE: DIFPRC'.0 ) GO TO 600 DIFRA(I) .0 PPX .0PFX)#(t.PFX*PFT V2* I'F X'PF T V2.NIE.
RFGR.I.2) 7 FORMAT( ItII .6F5. 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 .OMT GO TO 200 500 WRITE(G.2) G FOIMAT( III .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 .OMT I.2X.20X.7) 5 FORMAT(9F5.RFGT.
O(N).0.0t4O01 X( I ) I~) 13( ) 0(i1) NM I *N I 00) 2 1I.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.SUarIouT INE TDM(A .D.NMI .I KN a (X(I() X(K) RE TURN END  X(K41)*COK))/O(K) VAR09Z350 VAflo9:160 VA1109370 VA11093001 VAI.C .C(N).o(N).1O9:190 VA1109"IO() VAn0094 t VA11O0d1O '4A110943?0 VAIIW14 40 VAfl0O450 VAI*0946) VAI?94 70) VAR0O9400 VAR094 90 VAIR095C VAflfO!'j 10 VA110IJ520 .X(N).XN) DIMENSION A(N) .P(400).
LE0.K2NC JKZ2 .0.UG.0 D'rF2( I .F2(J.) G0 TOrOOiC D'. DA'' G.L!S( (]KZZ) ) . C C C C U AND v LDE VELOCITY EDU. iB10 .5(V( K)*:(.2D LUGRDX/(AUGOX +~DG'RD)' .ItD.3.VGRD2/(.G.TIONS SECT ION AD !ST ]F(ICLL].S .]K ...)K2NC " (Ki)'NWPC .E230 .FX(3. I O50 VI8090 K1 K12 2 KEP CONSTA'NT CLLCULATION C CRO5SFLD'.S(0.C "hT. 1CONV. D2).UG.K) z 0.:X GO TO 40000 "0OO D.LT.0 OOC COI~'T2]'UE C VAR P182 vLR a210 VAR E220 VAR. I vtR I5S0 rooOT 00G2 UC 'S(U(]K) ) VLRi 1210 vAR1E230 VAR E20 U(3KXX) ) LUGc .xx'..DFFUSI'O =..K) a LUGDX.E1O 0005 FP" * IF(.E0.LT. DNF.EO.I.G  V RIS150 VAR.P2) GO TO 40002 (U(iK)*U(KX))) LUG = LS(O..EO.KS 2.K) * 0.PM.NPC o1 ]F(K. 1) GO TO 400 IF(K.CR.I.C .1) GO T0 49998 SOLUTiON FOR SIE.L.O.FFX ( .K) • AVG.KBP2 " (3.K t KXX a jK .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.P..S( 4 .JSP2 DO O40000 DO 4000 K.E0.KZ2))) .
I ) I KIPC # (KI)'tN'PCL VAR16D VAR 11510 VARi18520 !PK a 31. 054 (K..]BR ( '.E. a uJD(2K) )0.0.I) K2WC' + II)K NC *K2NC "00Z0100 K=2.5 UODUMKS) 16D(1PK) * .'O('. . UD( 3K) UL UR UT Uo PK) UD( 3KP) U0( 3KMS) DCR DCL D!FFCD(IKL) D) F CO( KL.PKKNUO(IPK) )..S(UftH)(UCUQ)  VL21S660.ET/TADD3 VAR IS420 VLO IS440 3F~lDw. ( '(K ULH C CON'. KS I KLS a3)K I KP a IK  NP 2 I N1PC VAR 185501 YLR 15560 VAR 1C570 VARias590 VAR 86201 VARi1Es30.E003 DO E0100O ) K GO TO50205 222.5 )r0.1.0.R2CLI. 3F(3~.. VLRIE6t0 V4RI15650 VAR 15690 ULWM'(uLLIC) ( IR'(L'C4UV) VARIE *7 DO.1 SRKSrI a * N BUOJY a LP0E7.2)GO a TO 29990VALR iS.5 PC) ').KSPI I KL I lK z lK 1A 0.C RD20 3KS0 S3G'~ to VAR 8340: VARIS5360' RDZDRD2 VAR i !:0.I DIFFCD(3KL2) D!~FFCD(IKL3) *0(!.P( TJ"'.
Cw'))'A.% At t x NdE ONCA + lA b + A t + OZ~N(Lt) OdAN 00gla aVA 06!.o2m.'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 )*.) ..8~ VA 0899tMVA (oaQN 3NC647.)10 6666a 01 0 MfN i NGO 07 6 "VA ZNOO NCO ZOM( + (Xt)0M £ ) O~t6tVA (~ ~ ..(onin) M0'a(On.=OStd7A 0506LdV  ~ ~ ~ ~ Hm )CC7r~! Sall*~ ZN07J ±.OSZ6 M7A OdI'N Al dN O7I~t 17A OdM ( A) ONZ Nt + 1IOZN4(Vt) * + 3NZ)(11) + t bi > >4 E A c TA .( (nf1l)(Himlss7 (In3n)(uLM)S9V ) ZNQ..+fl)IdM Y(Qd.onH.(?nOn) i13( fl2f) unc.(. (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'.t) 2 td9N'Zt oo0to oa 3flN tLNOD s0?_O3 M 6 dVA + NAcns + ZntO * (Nt)d(dXt)d ).7A 0060tMIA 0L6Sgv7A 0968LVA I 3a CA~ SimI) 0L96 t'7A O9S63tM7A j0 0 VA d061Nt) AoflsaNx0As + Od.(iim1 (.o.uQ + + t (Nt)0ftl x = XIt)f ( oo.vt Z1t0 7 Z(J~Si0(o(nsn)(a)a (1n.
DIFFCO(IKL) DCL a D0FFCD(IKL+2) URT x 0.( DCL * DCR ) C1(J) z 0.5 AI(J) a 0.VRE + ABS(WRB) 1 * SIGMA*RDZD.( DCT + DCS ) C1(J) a 0. = DIFFCD(IKL+3) WRT 3 (W(3K) + W(iPK))O.SIGIArRDZD3DCT 01(J) a U(IK)ROT 81000 CONTINUE C .K2NC .DI.IKBR) CALL TDM(A 1.ABS(VRT) ) .. VAR19310 VAR19320 VAR19330 VLR193AO VAR19350 VAR19360 VAR1S370 VARie380 IKL a C 1 + (1)'K2NCL + (K1)NWPCL IPK .IBR DO 81100 K2.( URT EAS(URT) ) .KBR) C DO 82100 Ka2.RDX.W(1K)RDT 82000 CONTINUE C CALL TDM(A .ULT * ABS(ULT) 81(J) .SIGMARDXD*DCL ACBS(URT) .C1.5.C1 .'5 WRS (W(1KMS) * W(IPKNWPC))O.5RDX( URT 1 + SIGMARDXD.ROT * 0.Ke1P ) VAR 19390 VAR19400 VAR 194 10 : VAR19420 VAR19430 VAR19440 VAR19450 VAR19460 VAR 19470 VAR19480 VAR19490 VAR19500 VR 19510 VAR19520 VAR 9530 VAR 19540 VAR19550 VAR19560 VAR19570 VAR19580 VARi9590 VAR19600 VAR19610 VAR19620 VAR19630 VAR19640 VAR 19650 VAR19660 VAR19670 VA2 1960O VAR 19690 VLR 19700 VAR19710O VAR19720 VARIC720 VAR197540 VARS19760 VR 197 70 VAR19780 VR 19790 VAR19800 .ABS(ULT) ) .AeS(WRB)) .NWPC DCT 3 DIFFCO(IKL+1) DC.SIGMhARDZDODCB E1(J) = RDT * 0.]K .X.]K + K2NC IKP a 1MKS 1K + NWPC ]K K K2NC NWPC 1KIAS = C DCR .SIGt.KER D00 82000 12.1 ]K £ + (1])K2NC + (Ki)NWPC ) VAR 19260 VAR 19270 VAR19 280 vLRlg2O VAR 19290 VAR19300.0.5( U(IK) + U(]KP) ) ULT C a 0.IMKS IKt.5*RD2( WRT .KSPI ]K = 1 + (11)K2NC + (KI)NWPC 1X a (12).KBR DO 62100 1=2.'RDXDDCR D01(J) .5RDZ*( VRT * ASS(wRT) .5*RDX( ULT .5RD02( WRe .X.$ ]K .KBR + K 81100 U(IK) a X(IX) C DO 82000 K=2.1SP J a (K2)]IeR * 1 .8e1.5( U()]KS) * U(IMKS+NWPC) ) Ai(J) = 0. C 00 l8100 I=2.16PI 1K a 1 * (]i)K2NC + (K1)NWPC IX x (K2)ISR + I 1 E2100 W(]K) * x(Ix) 00 823000 K2.
L 0.' ).v%4!)iS +.40 C0~7A O CO0d7A NAoC1s (A(I)a .(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.Htin 0~Q0~*/'.3nwNtJ.fl.Q ( (S'feXtiM 094=J7A OSZO0t7A (>I).LG!1CE)PA HIM. 3)3QO.3 00 000O~d7A "L661 aVA (AI  (Ad 1) d !OSGOZi VA OG61 !VA 03661 SVA b33aXac~tjS ( 100 (Hll)SV H~ln   C(Hzn)sev  Ya   ~ A +Laa Hfl )xCi~s~o = x (r)La ()9 t ()1 2 (H. z (r 01)~S~7~).'%)SS'7 + HSA .~tSC(Hifl)ss b3(3 ).mNtAz + + ONA(lt.S'Q g* 2 S = CI)' 1 C + ># O oc OZMVA ~dN OdMN )Z' 1I A W> = S AVI + A!t d>'[ 06t0MIA M17A 0910~1VA 0C10aVA +N *) 1:)dfMN'()) + IONZ)A(Ll) OdAIY( .ZC(3aE1~S (d~t)d  00oo' ((HIiA)SS7 I )'ZQZ .)ssv .v4)tS ~m * io1. ).0xCd. )'ZQdS50 )Zc!S*  a ' a  (HS.~) Al NP 1=I000ca 00 .(HL')S..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 ).
NWPC CFR = CF(1PK) CFL .EO.12 J .B.2.S]EO(]K) TSRA TSLA .D1.NSEO) 33333 CONTINUE 15E0 a 0 DO 60000 K=Ki.S) SIECO .EO.CF(UK) IF (CFC. IBR(K2) + I ' (1i)K2NC IK = i + (K1)NWPC IKL a 1 + (11)K2NCL + (K1)NWPCL RC a FLOAT(I1)bDX .I) GO TO 55000 DCR .Ci.1) GO TO 2040 ISEO = ISEO + 1 IF(ISEO.1CALI.25*RRC CYRX = CYLO0.U(JMKS) vW = W(IKMS) AUC .1SP1 00 84100 K2.1 (12)(KER1) 84100 W(JK) a X(lX) C 89999 CONTINUE IF(ICONV.KSR IK IX a I + (I1)*K2NC + (Ki)*NWPC + K . VAR20420.LE.0.DLeS(UL) ALC = DAES(WC) 1wS * DASS(WE) = SIE(1K) S!EC SIER .SIE(IKP) S]EB * SIE(]KW.DIFFCO(1KL*i) DCL a D!FFCO(1KL+2) DCB * D]FFCO(JKL*3) ]PK )KP a a + K2NC * NWPC = 1K .IKBR) DO 84100 l=2.K2NC ]K IK 7]KS IKMS = IK .NE.5RRC*RDX CFC .DASS(UC) AUL .CALL TDM(A .X.DIFFCC(IKL) DCT .HDX RRC a I.SIE(IMKS) SIET .5(TS(]K)+TS(lMKS)) VAR203G0 VAR20370 VAR20380 VAR20390 VAR20400 VAR20410 .O.CF(]MKS) UC c U(IK) WC = W(!K) UL .5(TS(IK)+TS(IPK)) . 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 .SIE(IPK) S*EL .K2 DO 60000 13a1.0/RC OR = CYL0.0P.
(N t0 *090~bVA 0O0I bVA 0O'0t MVA OSOI aVA 0OP0tZSVA (Q QtlY.OSrlth7A OV7 t!VA t."ai 0*0 000 09ML VA 3flNIINOD 00009 =(fil) 0O~VA 0Lt t aVA OOtZVA 06tt~7A XC(QXC + dAO)dQZJlj + 01 0 00009 02. C0 r~t (r)LC 00~ttIdVA 060k ~wVA 0OLt dVA 00009 01 13t5.± 4:7) 21J3 1u L7S1L'q'l OS 60ZMVA (Slx () I s I)SO I ((diE)l+ m1)l)'c = vzs I VIL . t :) a . N *:dM o zi Z)q4s)4 coo 0. 10 + ( 0S669 01 00 (V03VIJfl S0.L.+c a 21J10 to.~t J '~1f ((t0AX'1ZJANO ) C dt0* ) * N ) S 4~ I .At~VA OSZI~7A clt'L fa ZAVtAZ'4 0=0i9 00 0=O9 00 I .(r'..x)j! 2(AI1)3rS oo0t 9 06CI ~1VA' 09ZI dVA 0LI. t7) .Q ~ 000 fl co(3 S6 .OlI)03(r v+n . 06C Zd7IA ZltrEzf 00 OOOL CO 0SCtLb7A NOI0t1D3JCZ NE d33MS CNZ 0L~tZZ7A LCECE OZ :tZ7A 1 05 3 3lNTI.1! Al z(IJ)L3 (NOfl'Ol V(ISJX017) Ot)t0dVA 0001ZlA N7 dlxiol ((L1)x~JNG4(IV~~sNQ)ts0 .C*~ t )4* . (nN*31.NOD (L03e.I vo 0009 0:9>4 I )c t a.
LE.i) GO TO 70000 IF(CFT.K)*DNFF2( .5"(ws*Aw )'RD2 .C2. .A2(J)SIE(3KMS) A2(J) • 0.5(DNFF2(3..0 GO TO 70000 65500 C2(j) E82(J) .0 E2(J) .0 DI(J) D(J) + DIFFTDCT*RDZ GO 70 70000 65000 01( ) = SIE(]K) L2(j) a 0.0 C2(J) = 0.DIFFCO(IKL) = D]IFFCO(IKL+1) = DIFFCO(]KL+2) .K.TTSPLL .K2NC IKWS VAR21460 VAR21470 VAR21480 VA021490 VAR21500 VAR215 O VAR21520 VAR2rS30 VLR21540 VAR21550 VAR21560 = ]K .1) GO TO 65500 D1(J) * 01(J) .NNOPT.(WC*AwCWS*iWB).0. 1 * 0.5(TS(]K)+TS(3KP)) TSeL .(12) + K .CF(IKt S) WS W(IKMS) WC a W(IK) VAR21570 VAR21580 VAR21590 VAR21600 VAR21610 AWB a DLeS(we) AWC DOASS(WC) TSTA a 0.CO(IK)0.K2 Ix = KS.5(WCWC).5(DNFFZ(I.12 DO 71000 KKI.5.(A2.EO.NU) GAMT * RPRAN .EO.X.DIFFTDCTRDZD C2(J) a 0.0.21730 VAR21740 A2(J) = B2(J) = VAR21750 VAR21760 VAR21770 VAR21780 VAR21790 VAR2800 VAR21810 VAR21820 VAR21830 VAR21840 VAR21850 VAR21860 VAR21870 VAR21880 VAR21890 VAR21900 VAR2190 VAR21920 VAR21030 VAR2190 VAR21950 VAR21960 VAZR21970 VAR21980 VAIR21990 VAR22000 .OlFFTDCTRDZD IF(CFB EO.E0.DIFFCO(IKL+3) 1PK a 1K + K2NC IKP a !K + NWPC IMKS a 1K .NE..RDZ C2(J) = 0.k)+DNFFZ(1.1) GO TO e9950 Dl(J) = 01() .IF C DCR OCT DCL DC8 C ( CFC.0 70000 CONTINUE C CALL TDM.5(rS(lK)+TS(lKMS)) C GAT DIFFT DIFFB C C * RDTSIE(]K) .IKBR) C DO 71000 ]=m1.I DI(J) a TGA.5 0.0.I) )XMX GAMTTSTA VAR21620 VAR21630 VAR21640 VAR21650 VAR216GO0 VAR21670 VAR21680 VAR21690 VAR2I700 VAR217 tO VAR21720 VA..NWPC C CFT = CF(IKP) CFB .01.RDZ .B2.1.11.DIFFEDCRD2D RDT * (DIFFTDCT+D)FFS*DC6)*A02D .0 69950 IF(CFT.1) GO TO 65000 .M * * IF(TS(]K).K+1))X.C2(J)SIE(IKP) C2(J) = 0.X GtA.ND.
K2 LWPC=LWPC+NWPC AT TIMEN ) .11.NWPC DO 2.5'( SIE(IK) + STOR(I.NwPC 71000 SIE(]K) = X(IX) C IF(KIRK.09 KzK1.K2 1K 1 + (]i)'K2NC + (KI)*NWPC 33338 SIE(K) = 0.E0.K) = SIE(IK) 333335 SIE(IK) .K) ) KIRK a 0 33329 CONTINUE C C NOTE.O) 00 33335 I11i.K2 IK v 1 + (1I1)K2NC + (K1)*NWPC STOR(I.NE. 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.12 DO 33335 K=KI.SIEO(IK) KIRK a i GO TO 33334 33337 CONTINUE DO 33338 1=1112 DO 33338 K=KI.1) GO TO 33333 GO TO 33339 IF(ISEQ.IK r 1 + (II)*K2NC + (K1). TRANSFERS VELOCITIES TO STORAGE ARRAY( 2040 KI1 K2xKBP2 LWPC=I .S)=W(KK+) TO(KMS)=TO(KK+1) VAR22470 VAR22480 VAR22490 VLR22500 VAR22510 VAR 22520 VAR22530 VAR225ZO VAR22550 . 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.
'inputn dat'. The file. 'inputd 3' and 'inputd 7'. The output The files.3(B).1 and Fig. Fig. 'input mom'.8. results are in Fig. is for calculating the natural convection flow field in a Cartesian closed compartment. 5. 5. 'inputd HUH' and 'inputd DM'. . 4. Fig. is the input to test the stability of the ADI scheme for momentum equation.6 and Fig. The files. 5. are for calculating the numerical solution by donor cell scheme in the problem geometry. 5.APPENDIX D TYPICAL INPUT FILES The file. The output results are in Fig.2. are for comparing Huh's And De Vahl Davis and Mallinson's correction formulas in the problem geometry.7. 5.
0 1.0 4.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 60.0 10.013 0.0 0.0 0.0 60.2 10.0 0.0 0.0 2 l.0 1.5 0.0 0.0 0.0 0.0 2 5 10.0 0.4 5 1 1.rTh FILE: INPUT MOM VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 4 4 O STABILITY TEST FOR MOMENTUM 2.0 0.0 0.0 4 0.0 0.0 10.0 1.0 0.0 2.0 1.0 0.0 0.0 0.0 1.0 0 2 1 3 O0 2 2 O O ADI 10.0 0.0 1.0 .9 0.0 0.4 0.0 0 0.01 0.5 6.0 0.0 1.045 0.0 1.0 0.t20 1.4 1.0 10.4 0.0 1.0 0.0 I 3 I 16G 1.0 0 0 1.75 10.120 1.44 0.0 60.7 0.0 0.01 0 1.0 0.0 1.09375 0.
0 677.0 1.90 0.5 1.0 1.0 I 43 1.0 NATURAL CONVECTION IN A CARTESIAN CLOSED COMPARTMENT 1.20 1.5 5 1.+20 60.2 1.0 6 6 50.4 0 0 0 0 O 0 0 6 2 0.0 .0 1.751.0 1.0 0.0 0.0 8 2 I 3 1 3 0 3 0 1 1.20 II I 1.0 0.20 0.0 0.0 0 0 2 1.0 1.0 1.0 0.0 0.0 1.20 1.5 0.20 030 1.0 10.7 1.0 60.0 0.FILE: INPUTN DAT A VM/SP CONVERSATIONAL MONITOR SYSTEM 'PAGE 001 5 10 0 1 0.0 32.0 0.0 0.21795 1.01 1.44 0.67 1 1I 2.01 0.0 0. 20 1.0 0.0 0 1.0 1.0 4.0 2.20 2 1.045 0.0 1.0 68.
0 0.0 1.0 5.3 1 I 1.997 7 1.457 1.207 0.0 9.0 9.497 0.409 1.01 0.0 2 1t 1 113 5.0 0.0 0.0 7.0 4 I 36 1.0 9.O.948 1.0 0.0 0.911 3 4 1.0 5.0 8 10.000 0.0 0.000 0 0 O O I.0 0.204 0.0 1.0 0.0 0.0 9.0 0.000 9 1.0 0.CASE 3 2.497 2.974 5 1.0 0.0 0.07107 0.70711 1.000024 0.0 0.47.061 1 0.0 0.0 1.0 GO.0 0.0 1.0 7.249 3 1.0 0.0 0.07107 0.09375 0.000 10 1.9 0.0 1.0 0.457 1.014 0.44 4.0 .0 1.FILE: INPUTD 3 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 10 10 0 0.063 ( ).0 0.991 6 1.045 1.0 1.0 0.5 0.000 1.OGI 2.000 10 1.5 1.0 1.000024 60.0 0.+20 I 0.0 1.063 0.013 2.0 0.0 0 2 5 3 0 7 6 1.0 1.0 1 .75 0.0 1.009 4 1.0 0.014 ).0 0.07107 1.0 9 10.0 1.0 0.05.0 0.0 0.0 1.0 0.0 1.II I 11 2 2 1.0 9.0 1.0 1.70711 0. 020 2.0 0.5 1.026 5 1.0 0.003 7 1.0 5.002 0.002 0.751 2 1.0 0.009 G 1.01 0 0.0 7.0 60.0 0.07107 111 3 1 1 2 5.0 9.511 1 1.0 2 1.0 0.0 7.204 0.940 1.02 1.207 0.0 0.999 1.0 1.0 7.0 1.07107 12 12 Ill 2 5.0 0.0 2 4 2 4 1.0 0.0 0 0 NUMERICAL blFFUSION .001 0 1.07107 7.0 0.O 2 1.0 1.0 12 11 2 121 t 5.07107 0.000 0.000 0 0 0 O .0 10.
626 1.0 2 2 0.003 17051 1.+20 2.0 1.17051 0.002 5 1.0 5.0 7.0 1.0 7.143 0.FILE: INPUTO 7 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 6 0.003 0.990 5 1.0 0.75 2.0 0.0 0.0 0.633 1.013 4 1.07107 711 8 8 2 5.0 1.001 0.0 0.0 0.0 0.07107 I 113 7 2 1.0.0 .0 2 7 8 5.0 0.333327 60.0 0.0 5.374 1.0 1.000 0 O 0 0 0 7 1 2 7 2 5.143 0.911 9.0 0.5 1.0 0.457 2.0 7.0 9.0 0.0 0.0 0.0 0.0 0.07107 0.0 0.0 I 1.0 0.0 0.01 2.0 0.0 60.0 1.0 10.014 0.44 4.0 0.000 6 1.907 3 1.07107 7.09375 .009 1.0 3 0.0 0.0 7.0 1.00.0 1.0 60.0 0.014 0.07 07 0.0 7.0 1.0 1.000 0 0 0 1.0 1.071107 0.00.0 1.045 1.0 4 2 4 1.0 0.167 1.0 5.0 1 9.0 1.9 0.0 0.333327 1.0 1.0 1.0 1.5 1.0 3.0 0.457 2.0 1.420 1.0 0.0 0 0 2 2 1 0.3 1 1 1.633 1.001 0.07107 1 1 713 2 5.0 0.4 8 2 1 3 0 0.01 0.0 NUMER ICAL DIFFUSION .0 4 I 36 1.01 0 0.47.0 0.013 2.0 1.0 4 1.0 0.0 2 0.0 10.0 0.CASE 7 0.0 2 9.000 6 1.4 1.
09 W '09 0. 1'0 0.1 1' 0*0 0 *1 D99C00 0'1 0IO' 0*0 0*0 0*0 0'0 0 *I0*0 9c 1 t.*I riI 00 0*0 0*0 0i'00 0 10*0 6 v 0*0 0'0 top' .L Z01 'L L'09C91C P09c91''? 1'Q9C9p 0'19C91' 1 iLO'O 1'LO*0 GL00O LLO'0 £00. 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 CGO.z 10'o 0 1 0'l GL'O 1O1 '0 I 0*0 01l 0'l L910* 199*0 099.u *0 Ul' I 6P I 0011* I 0.0 GLO10 ZU10 SL9*0 109.0 01l 0'00' C0'I C.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 . 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 .1bl b 690*6 £00*0 LC9 0 90c.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 9?LL '0 G'.SAS 3~~~)Vd kOIINOW 0 0'0 0*0 0. 0 LG0 ..0 0.0 09G"0 9Cs£0 661''0 6sv1'.
0 0 2 1.20 I13 1.20 2 1.6603 0.20 2 1.0 I 1 5.0 0.20 913 1.20 10 1.FILE: INPUTO IIU4U VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 2 0.6603 0.0 0.0 5.0 0.0 2 25 5.0 26 26 5.0 2 25 25 5.20 9 1 1.20 5.20 1011 1.20 1 1.0 5.6603 0.20 911 1.0 8.0 .
01 0 0.093 0.0 1.0 0.0 0.076 0.0 1.002 2.021 0.013 2.0 0.951 0.0 0.051 0.463604 2.133 10.0 1.005 0.GOG 2 1.02.360 0.932 1.0.030 0.OGI 0.463704 9.0 0.0 0.0267 0.003660 60.0 0.0 0.0 1.0 0.045 1.0 1.0 0.0 1.0036G0 0.242 7 1.365 9.30i9 1 0.144 1.044 9.0 1.0 0 0 NUMEtICAL IrFFUSION .0 1.0 3 10 1.0 1.002 0.0 1.0 1.0 0.0 0 2 1 3 0 7 6 I 0.4GI6 0.636 0.09375 0.700 .705 0.0 0..4 4 1 36 1.rILE: INPUIO OM A VM/SP CONIVEIlSATIONAL MONI(fro SYSTEM PAGE 001 24 f( 0 I 0.030 0.0 0.0 0.0 0.920 0..OM 0.002 0.460 4 1.0 0.0411 0.0 0.223 3 1.099 0.0 1.0 0.174 1.0 1.4 0.0 0.302 0.007 0.75 1.000 0.44 4.0 5.156 0.0 0.0 2 4 2 4 1.009 0.0 1.120 2.20 1.496 0.0 0.0 0..522 0.403604 2.115 0.566 0.0 1.3 1.0 1.0 0.0 0. 1.016 0.0 0.0 0.197 0.029 0.003 0.526 3 4 5 G 2.0 1.0 0.0 1.0 1.0 I 1.0 1.33G 6 1.0 5 1.0 0.5 1.0 0.0 1.0 1.0 2 2 2 0.034 0.0 1.969 0.5 1.072 ' II 12 13 14 15 16 17 10 19 20 21 22 23 24 0 0 0 I '.246 0.02 2.026 0. 4G1IG04 7.011" 0.0 1.0 60.02.0 0.767 0.003 0.5 1.420 0.0 5.903 0.0 1.260 0.0 0.0 1.2406 1.0 0.01 0.0 1.463G04 2.463604 2.0 1.0 60.003 0.9 0.96G 0.949 0.567 9.
GG03 0.0 0.0 2 5.SILr.0 .0 5.20 113 1.0 2 5.0 5.20 911 I.0 I 5. 6603 0.6603 0.20 I 2 1.0 0.0 26 5.20 5.20 913 1.0 26 2 1.0 0 29 25 I 25 2 1.20 1011 1.10 OM A VM/SP CONVERSATIONAL MONITOnr SYSTEM PAGE 002 2 0.0 0.: INP. & .20 1.0 I*  .20 9 1 1.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.