This action might not be possible to undo. Are you sure you want to continue?
https://www.scribd.com/doc/89039116/NumericalDiffusion
10/25/2014
text
original
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
3.2 Governing Equations 2.3.1 Mesh Point Implementation Mesh Interface Implementation 4.2.2 Truncation Error Diffusion 4.2 Limitations 2.2 ke Model 3.2. Implicit and 4.2 Turbuletit Diffusion 3.1 Assumptions 2.1.3.2 Crossflow Diffusion 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 2.1 Continuity and Mass Diffusion Equations 2.4 4.4.2 Energy Equation 2.4.3 Truncation Error Diffusion in a Two Dimensional Problem 4.1 Derivation of Turbulence Equations 3.3 Basic Assumptions and Limitations 2.2.4.1 Problem Statement 2.2 Comparison of Explicit.1 4.4.2 4.1 Molecular Diffusion 3.4.1 Truncation Error Diffusion in a One Dimensional Problem Another Approach for a One Dimensional 4.1.1 4.3 Tensor Viscosity Method Corrective Scheme by Huh 4.1 Skew Differencing by Huh 4.2.2 Physical Constraint 4.3 Discussion of Skew Differencing and Corrective Schemes Raithby's Scheme 4.1.3.3.2.4 Solution Scheme 2.2 .5 SMAC Scheme and Compressibility CHAPTER 3: DIFFUSION MODELLING 3.3.2.4 ADI Schemes Conservation 4.4.3.3.1 Diffusion Convection 4.3 Low Reynolds Number Flow CHAPTER 4: NUMERICAL DIFFUSION Truncation Error Diffusion 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
15 . .9 5. numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig..3(B) .. 5.8 . . . .10 Comparison of the numerical solution with donor cell differencing of convection term and analytic solution with increased diffusion constant.3(B) . .17 Comparison of the analytic solution. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure .13 139 5..11 Comparison of the true solution and solutions by donor cell scheme. ..16 143 5.. 5. . . . problem with 6 = 63. .. . .. .14 140 5.430 . .3(B) . by De Vahl Davis and Mallinson's formula along the line BD in Fig. . . .convection problem with 6 = 450 Comparison of the true solution and solutions by donor cell scheme. Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 6x6 meshes along the line . . 5. True solution and donor cell solutions with and without diffusion correction for a pure convection . 136 5. .. . . . 137 5. Donor cell solutions with various diffusion corrections for a pure convection problem with 8 = 450 . . 5. 144 . . . Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 10x10 meshes along the line CA in Fig. . CA in Fig. .12 . . . skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with 6 = 63.43* . . (D+DDM). . 142 5. 135 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 . .20 . 153 ..19 .10 5. ..19 . . .. ..... .. . .22 150 5. . Comparison of the true solution..3(B) ... . Diffusion constant corrections at each interface for the mesh point and mesh interface implementations of the corrective scheme for recirculating flow field given in Fig. . .. 5.. 152 5. . . . . . . .... . . . 145 5. 5. . . 148 5. . . . numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig. ... Recirculating flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme Detailed flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and riesh interface implementations of the corrective scheme Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .23 .. . 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. . . . 146 5.21 S. 151 5..147 5. Simple recirculating flow field for test of the implementation strategies of the corrective scheme .. . . 5.19 ... ... . . .18 Comparison of the analytic solution.. ..24 .. .
30 with . . . . . . .. 5. .30 5. .. . 0. .3(B) = e. 1 178 ..3 sec. .0 sec .1I 5.. .. Problem geometry for comparison of the explicit and ADI schemes .. Steam concentration distribution after one time step.3 sec. 156 5. . .. .5 sec and the Courant limit of 1. . 157 159 5.30 for At = 0.32 161 5.31 1i60 5. 155 5..cx T where 4 is given by.. .29 Steam concentration distribution after one time step. . . .27 . ..3 sec... 163 A. . . . Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. . with mesh interface implementation of the . 5. . with mesh point implementation of the corrective scheme .33 Comparison of the ADI solutions for the profile of 4 at t = 10. 5.0 sec and the Courant limit of 1. . .30 for At = 1. . . .0 sec . .. . . . . . different time step sizes Boundary conditions in terms of 4 for the problem in Fig.. . . with donor cell differencing of convection term without any diffusion correction .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 . . corrective scheme . . 5. 154 5. . . . . ..28 . 0.0 sec along the line AB in Fig. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. 0.. Steam concentration distribution after one time step. . .
. Comparison of the Von Neumann analysis and numerical experiments for the stability condition in terms of the cell Reynolds number in an explicit. . . .2 . 2D. . . . . . . . . . . . . 1981 . . . . . . donor cell differencing of the convection term . . . .1 List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research. . .12 LIST OF TABLES Number 4. Page .114 . . 78 4.
cp/c Length Mass Molecular weight Nusselt number Peclet number Effective Peclet number (1) Pressure u u (2) L* ax or Ax Ay Ax + p v C .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./Imaginary (1) Turbulent kinetic energy (2) Thermal conductivity (3) Ratio of specific heats.D y.d xy e f Gr h I Im k g M Nu P Pe p y Az Pr Q Prandtl number Heat source .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.C x y D D ah D as Dhs D .D d .
tan.1 v u Mesh configuration.14 Letter R Re Re Rt Rx S Sc T t At u v v w Ax y Ay Az Greek Definition Universal gas constant Reynolds number Real Turbulence Reynolds number Reynolds number in x direction Source Schmidt number Temperature (1) Time (2) Parameter of a quadratic equation in Von Neumann analysis Time step size xdirection velocity ydirection velocity Velocity vector zdirection velocity xdirection mesh spacing Mole fraction in multicomponent gas ydirection mesh spacing zdirection mesh spacing a ala2 8 6ij C p v 4(1) 6 61 Thermal diffusivity or diffusion constant Parameters used in Von Neumann analysis Thermal volumetric expansion coefficient Kronecker delta Turbulent dissipation rate Density Kinematic viscosity Vpiscosity Phase change (2) Any general conserved quantity Velocity direction. 1 tan STime a Z Superscript (1) Vorticity (2) Amplification factor in Von Neunann analysis constant of heat transfer Traction tensor due to viscosity Summation Time step n .
15 Subscript cell cf DM f g i=l 2 3 4 L t m mix new ob old R t total US Cell Crossflow diffusion De Vahl Davis and Mallinson's Fluid Gas Air Hydrogen Steam Liquid Droplets Left hand side Liquid Molecular Multicomponent gas mixture New time step Obstacle Old time step Right hand side Turbulent Total calculation domain Upstream .
the fast blowdown stage and the slow mixing stage. hydrogen. The governing conservation equations are decoupled in order to simplify the solution procedure.__ 16 CHAPTER 1 INTRODUCTION One of the major concerns in the Three Mile Island (TMI) accident was hydrogen gas accumulation in the containment. although it may be less than 100% when there is no liquid component present. steam and liquid droplets are assumed to be in thermodynamic equilibrium and the relative humidity is assumed to be 100%. The distinct feature of the second stage is a much longer time scale in comparison with the first blowdown stage. the response of the containment during an accident can be divided into two stages.I~~U1i* 11*11044~1 1~. Some remedies have been proposed to deal with However. From the hydrogen transport point of view. Some simplifying assumptions are made concerning the thermodynamic state in the slow mixing stage. The research work reported here is primarily con cerned with the formulation and validation of the physical models and numerical schemes in the second slow mixing stage. the fluid dynamic phenomena in the containment in order to justify those remedies. air. it is necessary to understand such problems. The four components.( 1~111111_~. The error due to decoupling is negligible in a slow transient where the .
Convection is assumed to occur as a homogeneous mixture. c. Convection and diffusion are the central issues in physical modelling efforts of hydrogen transport. Diffusion occurs by two independent mechanisms. and turbulent dissipation rate. of each component in multicomponent gas depends on the mole fraction of that component. The turbulent kinetic energy. The turbulent diffusion constant is predicted by the kc model [45).17 state change over one time step is small. In the first step the continuity and momentum equations are decoupled from the energy and other scalar transport equations and solved by the Simplified Marker and Cell (SMAC) method in order to obtain the flow field. . The molecular diffusion constant is predicted by Wilke's formula [73) for multicomponent diffusion and ChapmanEnskog formula [5) for binary diffusion. and the total diffusion constant is the sum of the diffusion constants of those two mechanisms. The diffusion constant. k. are determined by their own transport equations. resulting in the same convection velocity for the four components. 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. molecular and turbulent. The turbulent kinematic viscosity. the phase change is taken into consideration to maintain 100% relative humidity. Finally.
Therefore. the major obstacle in accurate numerical modelling of convection and diffusion has been the crossflow diffusion error which arises in donor cell treatment of the convection term in multidimensional problems. Since the error usually appears as an additional diffusion.18 which is the diffusion constant for momentum transport can be calculated directly from k and E. The numerical diffusion has two different sources. The turbulent Prandtl and Schmidt numbers are assumed to be equal to one. numerical diffusion. however the latter turns out to be the dominant error source in most convection dominant problems. truncation error diffusion and crossflow diffusion. Truncation error diffusion is a one dimensional profile error and crossflow diffusion is a multidimensional operator error [68) of the finite difference equation. The leading issue in numerical modelling of convection and diffusion is to minimize the error that occurs in the numerical solution procedures. has been used to describe the numerical error in general. Truncation error diffusion occurs in the flow direction while crossflow diffusion occurs primarily in the direction normal to the flow. The effective diffusion constants of the two errors are of the same order of magnitude. There has been much debate on the numerical diffusion and many schemes have been suggested for the past two . the term. This is because the gradient of the scalar quantity under consideration is small in the flow direction in comparison with that in the direction normal to the flow.
ADI and implicit schemes. have been tested for recirculating flow problems. This situation is partly due to misunder standing of the sources of numerical diffusion and also partly due to use of inappropriate approaches for its elimination. skew differencing and corrective schemes. The schemes presently being used can be divided into two categories. mesh point and mesh interface implementations. The conservative property is essential in a long transient problem like the hydrogen transport in the containment. although with different stability conditions. The corrective scheme can be implemented with any of the explicit. 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.19 decades. The mesh interface . Chang's methods [16) are examples of skew differencing schemes and Huh's corrective scheme and tensor viscosity method [25] are examples of corrective schemes. The corrective scheme is inherently better than the skew differencing scheme in that it is conservative and does not affect the simple solution procedure. 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. Raithby's [53) and S. However. Two implementation strategies for the corrective scheme.
The suggested physical models and numerical schemes have been used to simulate the LOCA experiments performed in BattelleFrankfurt [44) and HEDL (Hanford Engineering Ref.20 implementation has always given physically reasonable solutions and may be used extensively for all diffusionconvection problems. [49) includes all the Development Laboratory) [6]. simulation results compared with experimental data. .
it is essential to understand the fluid dynamic phenomena in the containment. installation of ignition devices.1 Problem Statement Once a lossofcoolant accident (LOCA) occurs in a Light Water Reactor (LWR). for dealing with this problem.g.21 CHAPTER 2 PROBLEM STATEMENT AND SOLUTION SCHEME 2. containment inerting. e. In addition to the pressure increase due to the primary coolant. After a while the slow mixing stage follows the initial blowdown stage and continues for an extended period of time... use of flame suppressants and enhanced venting capability. the hydrogen has received much attention in the safety analysis of nuclear power plants. hydrogen generation gave a serious concern about the integrity of the containment in the TMI2 accident. There have been some mitigation procedures suggested. Thereafter. The major safety concern is that of keeping the containment pressure below a certain level to prevent a large scale leakage of radioactive materials. a large amount of steam and water will come into the containment in the fast blowdown stage increasing the containment pressure. . In order to justify the design of any mitigation system. 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.
laminar and turbulent flows. e.. phase change and heat transfer between the gas components and wall. thermal nonequilibrium._ __ _ __ _ I_ ~ 22 Since a numerical method is suitable for this purpose.g. a good computational tool has been required to predict the hydrogen concentration distribution. etc. Therefore. some simplifying assumptions should be made to use the numerical procedure without impairing the acceptable solution accuracy. The most difficult aspects of this analysis are the complicated geometry and chaotic postLOCA conditions. .
6) Equation of state: p = f p (ii .T) (2.5) Turbulence dissipation rate: D 1 _ 1t StE xk Dt P Txk 0E ] + C t*( 1 p k ui xk ouk au.23 2.4) Turbulence kinetic energy: Dk 1 t 3k 3 P 1 k ok x .Vp i + 4.1) Momentum: pgI + d + V0 (2.2) Energy: p[e e + Vvh] + = VkVT + e fg (2. I+ 3x 1 .3) Mas s diffusion: S+ V*vp i .xk k 2 v t BT C2 + kg k Pr 3x 2k k t (2.2 Governing Equations The governing conservation equations are set up to describe the postLOCA fluid dynamics in the containment. (2. 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.t = VD. Continuity: V'v= 0 pl + Vvv ] = Vp + Vv = 0 (2.7) .
The subscript i denotes the four components as .15) Dt = v t (2.14) m = fk (PiT) (2. .t (2.8) f (v) (2.13) D.18) p = Epi (2.17) t = C k 2/C E (2. Im + Dit it (2.12) (2.10) v = vm + Vt (2.T) (2.T) (2. = f (Pi.19) Note that 1.9) D. = D.k + k t m Dim = fD(Pi'T) (2.16) kt = .11) k = .24 Constitutive equations: 'N = f (p .
u. The mass diffusion equations are summed up for the four components.i (4). one energy.20) i = 2 = 0 43 + 04 = 0 (2. 2. (2. air. w. three momentum. 3.22) There are eleven conservation equations. .k C for a three dimensional case. one continuity.1 Continuity and Mass Diffusion Equations The continuity equation is redundant with the four mass diffusion equations and there is'an inconsistency between them. liquid droplet. p. there is one moreequation than is required. i=14.2.25 follows: i=l i=2 i=3 i=4 2. Therefore. four mass diffusion and two turbulence equations with ten primary unknowns. v.21) D4 = 0 (2. The phase change occurs between steam and liquid droplets and the diffusion constant of liquid droplets is equal to zero. T. H2' steam. The turbulence equations are written in terms of the Cartesian components in tensor notation.
1 and Eq.25) Vp i = 0 (2.27.27 remain valid throughout the transient. at p= 0 (2.) i + E 1 (2.4 are consistent only if the temporal and spatial variations of the total density are very small. Eq.Vp. 2.26 can be reduced to the following Eq.26 . it is reasonable to assume that Eq. ap at S+ Vp = E(VDiVp i ) 1 (2.26) It is a reasonable assumption in a turbulent flow regime that the four components have the same diffusion constant. 2. 2.27) Consequently Eq.24) Now it can be seen that the continuity equation is valid if and only if the following Eq.25 and Eq. 2.) + V(Ep Jt i = (V'D.26 are satisfied.23) The sum of the phase change terms is equal to zero. Therefore. 2. . 2.25 and Eq.(Ep i ) a . Since the concern is only for the slow mixing stage after a LOCA. 2. 2. Vp = 0 (2.
2. .29 denotes the energy change due to the concentration change in a given control volume.3. = VkVT (2.2 Energy Equation The energy conservation equation. I at t + V [vp. E it p + e.29) The second term on the left hand side of Eq. 2.ih i ] = V*kVT (2. 2.30 is substituted in Eq. Eq.i = V'kVT (2.27 2. LeaPi LeQ.2.ih. p + £e.29 to give the following.31 so that the internal energy and enthalpy can be written in terms of the temperature. (p. Be.30) Eq. is also of an approximate nature and the exact form is given in the following.31) The thermal equilibrium assumption is incorporated in Eq. It is clear that the major contribution to this term will come from the latent heat of the phase change. ae.iei) + V'[Ep. 2. [c pihi . (2.2.i + V.28) The temporal term can be divided as follows.
3 and Eq. It is necessary to clarify these assumptions and their limitations for a safe use of the given physical models and solution scheme. 2. = c p1 .32) where ei i = eg4g + e 4) Sefg g e.c .32 become identical by defining average internal energy and enthalpy as follows.34) 2. P (2 33) (2.T h.33) h (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.T P P e Piei p ih p.1 = c Vl. Pic vi T c p p.28 (Picvi )~ +V"[vTZpic pi] + efg 9kVT = (2. 2.T 1 Eq.
3. vt. The four components all have the same convection The liquid droplets move with the gas components velocity.1 Assumptions 1. 2. The relative humidity in the containment is 100%. with no slip. The change of the diffusion constants.2 Limitations 1.3.29 2. The relative humidity may be less than 100% when there is no liquid component left. 5. 4. Dt and k t over one time step should be small because the turbulence equations are decoupled from the velocity field calculations. The four components are treated as being nearly incompressible and the phase change rate is moderately small. 3. The four components are in a state of thermodynamic equilibrium. The phase change rate should be moderately small because the solution scheme decouples the phase change term from the energy conservation equation. 3. 2. The temporal and spatial variations of the total density should be negligibly small because of the incompressibility assumption. 2. The turbulent Prandtl and Schmidt numbers are equal to one. .
Pi(4). the obtained velocity field is substituted in the scalar transport equations. T. energy equation and two turbulence equations.7.7 are a coupled set of equations with primary unknowns. continuity/momentum equations set and other scalar transport equations set. 2.30 2.5 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. v. In the third step phase change is taken into consideration to maintain 100% relative humidity. The SMAC scheme is In the second step.4 Solution Scheme Eq. the relative humidity may be less than 100%. 2. is used to update the reference pressure and reference density which is important in calculating the buoyancy force. Eq. 2. the conservation equations are decoupled into two sets. In the first step the SMAC scheme is used to get the convergent velocity field with zero divergence for every mesh. The equation of state. 2.i Since it is too difficult to solve the whole system of the equations and obtain a consistent solution for all the unknowns. .5.1Eq. When there is no liquid component left. explained in detail in section 2. It solves the . which are the four mass diffusion equations. k. p. . E. The phase change is assumed to be zero in this step. u.
which can arise with net inflow or outflow boundary conditions.37): .37)y (2. it is possible to modify the SMAC scheme to accommodate a slight compressibility effect.e.36 and Eq. i. the SMAC scheme can be used only for incompressible flow calculations.36) .37 as follows.35) xdirection momentum: au au + u+v at ax au ay 1 p ax + Pgx + a 2u 2 ax 2 2au + 2 2 (2. ay (Eq. Since the continuity equation used in the SMAC is an incompressible form. 2. However. the divergence of the velocity field vanishes everywhere.ax 2.36) ydirection momentum: av av ua av 1 + pg + a y x2 Nv 2 (2.. 2. Continuity: au ax av ay 0 (2.37) Py The vorticity conservation equation is derived from Eq. (Eq.1_ _~ 31 continuity and momentum equations to get the velocity and pressure field in an iterative way. 2. The continuity and momentum equations are given in two dimensional Cartesian coordinates in the following.
37 as follows. . 2.x ay ax 2 ay2 where au av (2. the socalled tilde phase and pressure iteration.+ y ax ay The basic idea of the SMAC scheme is to separate each calculational cycle into two parts.40) S (2. 2. In the tilde phase the approximate velocity field at time step (n+l) is obtained from Eq. 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. (Eq.37): a2 = 2 ax 22 axay a 2 ay 2 (a2D + a2D aD +1 2 at Re ax 2 ay (2.37 as follows.38) =ay ax The Poisson equation for the pressure field is also derived from Eq. 2.36) + ~(Eq.32 a C 2 a2 + u+ V = v+ ) at .36 and Eq.36 and Eq.39) where av au D= . ax 2 2. ay 2.41) where fn and gn are the explicit quantities at time step n.
the following relations hold after the tilde phase. 2. The impor tant point is that un + l and vn+l calculated from Eq.42) av . _ Pi+j2P +P iPij+ pij f(uv)  Dn + l D n D D Ax 2y t (2. 2. Eq.45) .38. follows. 2. au ay au av ax  (2. Therefore.39 may be rewritten as A finite difference form of Eq. 2.33 Since the pressure field in Eq.40 and Eq. 2.43) ax ay It can be seen from Eq. V2p = f(u.44 is given in the following.O 0 (2. a guessed pressure field or the pressure field at time step n is used to start the iteration. The equation of vorticity transport.44) Eq.41 are still unknown. shows that the vorticity is independent of the pressure field. 2. 2.39 that the pressure field is not right due to the fact that the divergence of the velocity field is not equal to zero. 2.41 have the right vorticity.v) D (2.40 and Eq.
2. The velocity correction formula is derived from Eq. no more correction will be made.48) The SMAC scheme can be extended to a slightly compressible fluid flow. 2.34 The pressure correction formula is obtained by conn+l only and putting D. The slight compressibility means that the NavierStokes equations may be solved safely with the incompressibility assumption. sidering the cell (i.36 and Eq. equal to zero.37 so that it does not change the vorticity implemented in the tilde phase.j) 26p( 1 Abx 6p = h + _) Ay 1 At( A + 1 Ay 1 Dt D (2.46) (2.49) . (6u) L =  pAx (6u) R = t6p pAx (2. Once the zero divergence velocity field is obtained.47) ) The pressure correction is given in terms of the residual divergence for every mesh.. while the continuity equation has the following compressible form. S+ at V"(p) = 0 (2.
Now the problem is how to converge to a predetermined nonzero divergence field.35 From Eq.uniform over the whole containment. This can be done with the following modification of the pressure correction formula of the SMAC scheme.49 the nonzero divergence may be derived as follows. The associated physical assumption in Eq.50 may be given from the previous time information and boundary conditions.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. 1 . : in n =Vp v  total net inflow of mass at the boundary during the time At. 2.50) The two terms on the right hand side of Eq. .51 and Eq. 2.p_ p at 1 min/V At n p n n (2.52) (2. V: total volume of the containment. D = V'v 1 p p at V p (2. 2.51) v p where m.
48. 2. The velocity correction formula remains the same as Eq.50.6p = . 2.(DD 1 1 + ) 2At( Ax Ay (2.53) where D O is given by Eq. .
As collisions occur more frequently. in the containment after a LOCA. 3. Although turbulent diffusion is greater than molecular diffusion by a few orders of magnitude. air. which should be modelled independently.1.37 CHAPTER 3 DIFFUSION MODELLING The hydrogen. 3. Diffusion occurs by two different mechanisms. components. the process Since there may be three gas of diffusion is also increased. 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. the diffusion constant of each component is calculated separately by Wi~k's formula [73] in Eq. hydrogen. steam and liquid components are transported by convection and diffusion in the containment.1 Molecular Diffusion Molecular diffusion occurs by the collisions of gas molecules. air and steam. Convection occurs as a homogeneous mixture resulting in the same convection velocity for the four components. molecular and turbulent. 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 .
.AB where DAB: AB Binary mixture diffusion constant in [cm /sec] . The property of air will be replaced The resulting diffusion con with that of nitrogen gas.2) poAB D.1) where D. stants are given as follows. 3.: yi: Diffusion constant of the ith component Mole fraction of the ith component D.0018583 1 + M A MA MB (3. Eq.: Binary mixture diffusion constant between the ith and jth components.2. 1Y A B DAB 1Y B yA DBA 1Y C YA CA C YC CB B YC DBC C D = C YB DCB (3.38 the binary mixture diffusion constants and mole fraction of that component. The binary mixture diffusion constants are calculated by the ChapmanEnskog formula. T3 D = 0.
2.UIIII ~ II~PI^ICIL1I n~s^  ^ 39 T: p: aAB: MAMB : Temperature in [OK] [atm) Total pressure in LennardJones parameter in [A] Molecular weight QD.3. ap = np £ (o n + o p) EE e (3. The error range of the ChapmanEnskog formula is reported to be about 610% [5).lll~*. 3. 3. In order to calculate the para meters aAB and 0DAB in Eq. For nonpolar gases OAB and ODAB are calcu lated by Eq. OAB = (o A + oB (3. =n [ p n = [+ S n .3) CAB = AB For the pair of polar and nonpolar gases a correction factor is introduced to account for the polarity. (3. P n ]. the following relations are required.5) 2 n ..4) np = np 2 where the correction factor 4 is given by.AB: Dimensionless function of the temperature and intermolecular potential field for one molecule of A and one of B.
6 x 10 The fitting functions for the binary mixture diffusion constants are obtained by the leastsquarefit method. H2 a n = 7.9 x 10 25 25 25 [cm3 3 [cm ] : N2 : an = 17. 3.9492 x 10 D ah D 1 T .6810 x 10 . t* .= 1. 5 = 4. 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.6947 as .6 T1.2 E/K = 380 0 K a = 2.65 A The polarizability is given for the hydrogen and nitrogen (air) components.40 The parameters in Eq. The containment pressure is assumed to be1 atm.5 are defined as follows.9035 = 4.
k. n mix = Yipi n i E j=l Yj ij (3.8916 x 10 where T: D: [OK1 [cm 2 /sec] 5 T 1.10) Viscosity in [poise) Temperature in [K]) LennardJones parameter in [A) Dimensionless nTumber .67 x 10 where : T: a: Svi : o2 2 (3.lllI_ 41 Dhs = 2.lii.__Li~n_~PIIII ~. = n i=l y.8) j=l where ij 8 Mj j ij M The viscosity of nonpolar gases may be obtained by the following [5).8144 (3.6) The viscosity and thermal conductivity can be considered as diffusion constants for momentum and energy transport and calculated similarly as follows [5). i (3.7) k. = 2.
c (3.11) for the Prandtl number of Pr = c (3.[5].42 The viscosity of steam is. seems to be the kc model originally developed by Launder and Spalding [45). The .009 cp at 1 atm. The ks model sets up the transport equations with some empirical constants for the turbulent kinetic energy. There are several models suggested for Presently the best model calculating turbulence effects.12) where c P is the specific heat per mole at constant pressure. 2 n Yi umix = C 2 i=l Yi + 1. The effective diffusion constant is therefore the sum of the molecular and turbulent diffusion constants. E. Wsteam = 0.385 nn y.y RT k=l PMiDik k i There is a useful relationship nonpolar polyatomic gases.2 Turbulent Diffusion The diffusion process is greatly enhanced by the turbulence of fluid flow. k. 3. 200 C There is another formula suggested for the viscosity of multicomponent gas in the following [5]. and turbulent dissipation rate.
u. u. 29. 45] The transport equations for the turbulent kinetic energy and turbulent dissipation rate are derived from the continuity.1 Derivation of Turbulence Equations [15. etc. momentum and energy equations by decomposing the variables into mean and fluctuating parts and averaging the resulting equations. 34. u.u.2.= v ( Dt + ui j u P= Dt ax x 2 k6 ij 3x_) 3 i (3.' etc.' For example.43 turbulent viscosity is given directly in terms of k and C. 3. . = u. These terms are treated by the eddy viscosity concept of Boussinesq and eddy diffusivity concept which are given in the following.' j T'= a t x. The effect of turbulence appears as the additional terms due to the product of fluctuations which do not necessarily cancel out. (3. + u.14) . + Pi Pi = Pi + p.13) (3.15) .
2.2 ke Model The transport equations for the turbulent kinetic energy and turbulent dissipation rate form the basis of the kE model.c +g t Pr v BT az Tt (rue) + + 1 r ar a(w)= z [) [ pr ar o ar + z a ( a az + C 1 aw [ 2 au 2 t + 2() 2 () p az ar p k 2 2 + 2 2 C + ( r + ) az 2 2 k + k Pr atT zt (3. t + at r a (ruk) + (wk) = [ p r (r oka r ) + (t ) az ok a au aw + ) r aw au 2t + 2( + 1[2() 3 3r p + 2 + u2 2 ] . 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. A two dimensional cylindrical coordinate form of the kE model is introduced in Eq.' u is transformed to the turbulent kinetic energy equation by the contraction.16.44 The transport equation for u. 3.16) . 3. j=i.
metry.92 ok 1. C 0.0 o 1. = C pk 2/ (3.09 C1 1.3 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.17) The suggested values of the constants are given in the following table.3 The molecular effects can be included in Eq. Since the kE model is applicable to fully . it ak t a +it m ok lit m a E C The turbulent Prandtl and Schmidt numbers are usually assumed to be equal to one.44 C2 1. 3.16 as follows.45 The turbulent viscosity pt is given in terms of k and E in Eq. 3. it is difficult to determine whether a region under consideration is in a turbulent or laminar flow.17. 3.
lit k + l ak aui i + t (ax k + xk auk axi axk e . 3.5/(l+Rt/50)] C2 =2 where R 2 0 [1. The following kc model is a modified form by Jones and Launder [40.3 exp(R2t ) (3.18) In the above equations C 1 1 ok and a retain the values assigned in the ordinary kc model.46 turbulent flow. 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. 41).0 a ui ( 2 p ax ax 2 ) Dk Dt 1 p ak [(. The model presently being used is just to use the sum of the molecular and turbulent contributions for every region in the containment. CP = CP0 exp[2.0 .0. while C1 and C2 are to vary with turbulence Reynolds number. Rt. Eq.16.19) = pk2/VE . D Dt p ax k [(t Dt+ ) e k + C k i + k x ) au axk C 2 E2 2 t 2.2vak 2() k 2 (3. it is required to extend the model to laminar flow regime or to switch to laminar models according to some criterion.
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)
a(?) ) 2(.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. is for a one dimensional case.6 from Eq. 4.x+a) and [y. n+g 4 i+ Ax n i Ax 1 xji+ n i+l . .a i = ( nn+g n+g ax i+a  < a < n i1 8a ( ) n i+b a 2 4n+g ax 2 i+c 2 ox n 1 < b < 0 n+g (a n+g ax i.i+c T i+d ax where the following relations have already been used. 4.y) + (f) af (4.8) 3 @n + u()n + a( ax i+b ax 4 n+g _ n )n+g . n .51 Subtracing Eq. 1 < c < n 2 4 + i Ax 2 11 4 2 1 i+d < d < 1 The mean value theorem.3. 4.7. It can be extended to a two dimensional form as follows. we obtain the error at time step (n+l).y+b]. 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. af f(x+a.y+b) = f(x.
10 are at some appropriate points in the domain [xi . .A fAxAt axat 2 B utx(a+b) .9 reduces Eq.D +aAx(cd)Atax 2 +agt .10 are based on the assumption of perfect information at time step n. C and E are truncation error diffusion terms while the terms B. E . The terms A. 2 n+l (4i ) EXACT  (4+) FD = i n+1 f )2 2fx 1f 2 ) (f 2 2 .F (4. 4.52 Now Eq. 4. 4. 4. Although the error terms in Eq.10) The derivatives with respect to time and space in Eq.8 to the following form.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.xi+ ] and [tn tn+).tax . 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.
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. 4.12) and C 1 =N cL 1 e C2 ecL 1 . 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. The analytical solution of Eq.53 4. 4= Cecx + C2 where c = u/a . cL (4.11) ax Consider a case in which the domain is divided into N equal meshes and the mesh size is Ax. 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. u ax a22 (4.1.11 is given as.
Equation 4.i + 4i2 Ax = a (4. i+  i = (P+l)(4)  4)i (4. (4.14) The parameter P in Eq.14 may be recast to Eq. . 4. i+l i = (P+l)( 1 .14 is the cell Reynolds number or cell Peclet number according to whether the diffusion constant a is the kinematic viscosity or thermal diffusivity.13) where the velocity u is positive. P(4). i u x i+l2.16 is easily obtained from Eq.16) . Equation 4. where P =u a  l i24 + ii (4. 4. 4.15) Equation 4.40) .54 Equation 4.15 and solved for in subsequent procedures.11 is finitely differenced using donor cell scheme for the convection term in the following.15.13 can also be solved with the given boundary conditions by reducing it to the following form.
'N1 (1+P) N1  ) 0 N . 4. 4. 2 1 = (1+P) (4 0) 4' N 92 = N2 (1+P) 2 (  0) = N. 4. 4.12 and numerical solution Eq.19) (1 .LI II . i  0 = 4.19 into Eq. as.17.1 Inserting Eq.(1+P) P (1+P) 0 + N .20) Now the analytical solution Eq. 4.I_.11 and Eq. 1 (1+P) i .1 (1+P) N 1 (4.16.~LI LIII~. Both the analytical and numerical .13. (1 0 ) is ) P( N ¢0 (1+P) given as follows. (1+P) 1 . we obtain the solution for 4.18) (l+P) + 1 Therefore.41 = N ) (1+P) [(1+P) N1  2) (4.17) 01 can be expressed in terms of 40 and 4N N N by Eq.~l*IIIIX1~~^ ~ ^I ^~Xi 55 Then.20 have been obtained without any approximation for Eq. 4. 4.
12 is substituted for 4i Eq.+1 i + ) il 0 (4. 4.24) It is also possible to go through the similar procedures with a central differencing form of the convection term in Eq. i+l 2Ax il S S2ax 4) i+l 24i + i 2 i (4. 4.25.13. 4. 24 j. 4.1 (4.23) (4. 4. Pe = in (1+P) (l+P) iIp2 N (1+P) . 2 ii P ( (1+P) (1+) N _ 1 N 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. The in analytical solution Eq. Eq. in Eq. and the effective diffusion constant for the numerical solution will be obtained.22) Comparing Eq. 4. 4.25) (4. e (il)Pe (ePe1)2 NPe .56 solutions will be reinserted to the finite difference equation. 4.13.25) . + 24 S i i+1 i1 =e e l NP  1 (e P P 1) 2) ) ( (e 0 (421) The finite difference solution Eq.13.1 e Therefore.20 is also substituted for 4.21 and Eq.
26) where 2 < P < 2 The effective diffusion constant De can be readily obtained from the following relation. 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. 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.26 may be a good indication of the truncation error diffusion occurring in a general one dimensional problem. The convection terms in . it helps understanding the origin of the truncation error diffusion. 4. Although this approach is approximate and overpredicts the truncation error diffusion.57 The effective Peclet number for the finite difference solution of Eq. Pe = kn (2P) 2+P (4. Therefore. 4. Eq. 4.24 and Eq.25 can be derived as follows. uLx D De Pe  (4.1. As in the previous section the velocity u is assumed to be positive and constant.27) In general.
5 may be approximated as follows. Therefore. 4.4 and Eq. the profile of 4 will be linear and the truncation error will be reduced to zero. the convection is dominant over the diffusion. 4.28) Ax (4.29 hold only when the Peclet number is large.29) i+ The truncation error diffusion can be quantified by subtracting Eq.( i+f 2 (4.58 Eq. It can be shown that as P goes to zero. If the Peclet number is small.30) Both Eq. Pe also goes to zero resulting in no diffusion error in Eq.26. 4.31) ND . the truncation error diffusion constant for a convection dominant problem may be given approximately as follows. i. 4.e.29 from Eq. 4.24 and Eq. 4.28. 4. the diffusion is dominant over the convection. u i Ax i1 u i+ a4 u() ax u(a) ax (4.e. 4. D uAX 2 (4.28 and Eq. ae u( x where 0 <f < ae + u() ax i+  u~x a2 ) .. i.
33) . the total diffusion quantity of 4 due to the truncation error is not significant in comparison with that due to the crossflow diffusion.3 Truncation Error Diffusion in a Two Dimensional Problem There are two sources of numerical diffusion in a two dimensional problem. 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. two dimensional. Since the gradient of 4 in the flow direction is negligible in a convection dominant problem.1. 2 u + v 2 + +( ) (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.59 4. one is the truncation error diffusion and the other is the crossflow diffusion.32) = x 3y Equation 4.32 is finitely differenced using donor cell scheme for the convection term. A steadystate. conservation equation with no source term is given in the following. u i 6x ilj + v i by ij_a(i+1 2 4x i + il1 +ij+l 2i 2 + 0iji by (4.
v( ay ) ij .n) ( = x cos6 + y sine I = x sine + y cose (4.x[cos + (cose tane 1 + sine) + sine tane ] ax xY (4.2 is applied to Eq. then the truncation error of the convection term is given as follows. transformed to the coordinate ( .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 = . 4.33. ij Ax 1 1x Axhy 'i1 i Ay iji 1 +AxAy fj+ j i+ i axx dt + Yj+ yj+ yj xi+4 xi+ a (v)dx dy ay Xiih + u(t) x i+ a =U(ax) )3x .34) where tane = Ay/Ax tane1 = v/u The coordinate (x.35) .y) is by rotation.60 The approach in section 4.1.j+ + v(a) ly i+h.
36) where U = /u2v 2 Now Eq.n) given by Eq. 4.37) Equation 4. the problem is basically one dimensional and the truncation .35. 2 2 2 + a = ( 2 ) (4. y v e Sx Then Eq. e is equal to 0 or 7/2.32 is reduced to the following form.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.34 will be transformed to the coordinate system (t. 4.61 where 5 is the coordinate in the flow direction. 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. 4.
62 error diffusion occurs in the flow direction as in a one dimensional problem. The crossflow diffusion comes from the multidimensionality of the problem.3.1. 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. tional false diffusion was explained by Patankar [52] and Stubley et al. It is entirely different Recently this addi from the truncation error diffusion. which shows a single mesh with pure convection. 4. 4. therefore it exists only in two or three dimensional problems when the flow direction is not aligned with the mesh configuration.68) and it was clarified that this is the dominant source of the error in most multidimensional problems.l1(A).1(A) is transformed into . The origin of the crossflow diffusion is illustrated in Fig. 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.2 Crossflow Diffusion Most of the confusion about the nature of numerical diffusion comes from the fact that there exists an additional source of false diffusion. that will be named here as the crossflow diffusion. 4. 4. In the donor cell scheme Fig. [67. In Fig.4.
4.1.63 Hot Mledium Hot 7 Hot Cold (A) Cold (B) Cold (C) Fig. Illustration of the crossflow diffusion in a single mesh with pure convection .
. Then homogeneous mixing occurs in the mesh. This is a pure convection problem Consider the fluid element The fluid with no physical diffusion. 2 (lp)[l+p+p + . diffusion occurs in multidimensional donor cell differencing of the convection term. 4. at those points is exactly equal to 4. an appreciable amount of false program user. 4.1(B) where the velocity components normal to the surface are considered. A 2 .&x u v ax Ay The effective diffusion constant for the crossflow diffusion will be calculated in the two dimensional Cartesian coordinate in Fig. 4. of the value of 4 causes the crossflow diffusion.3. . 4. ] = (lp)4 = where u .1(B) are again interpreted as Fig. .2 which shows a row of meshes aligned in the xdirection. A 3 . and the sum of the values This distribution along the points A1 . in The numerical results Fig. . CAB which looks like a long onedimensional rod. 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. In the donor cell scheme it is distributed all . Another illustration of the crossflow diffusion is given in Fig.64 Fig.and intermediate temperature fluid will come out of the top and right surfaces. 4.
65 (1p)4' x p (1p) 2 p (lp) p3 (lp)4 p4 (1p) A5 Ay A Ax 2 4 y p t* U where p = Ax Ay Fig. 4.2. Illustration of the crossflow diffusion in a row of meshes aligned in the xdirection with pure convection .
66 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. V A2 e A.3. 4. Geometry for explanation of the crossflow diffusion in a two dimensional Cartesian coordinate .
Diffusion A t1 . k2 = Ax sine/sin(el+8) k3 = Ax sine 1 /sin(e 1 +8) After substituting appropriate expression for each term. right and left hand sides. 4. v.4. sine cose cos= sx Uacos61 sin(O+e ) I DRHSS .67 element was on the line CAB at time t in the velocity direction. Gradient) A /k 1 =A of 4 Current 2 2U of 4 Therefore. the following right hand side diffusion constant results. which shows that the value 4A diffuses out along the fluid element. The travel of the fluid element is analyzed in Fig. and is translated The value 4A at the point A should reappear at the Point A'. In the donor cell scheme it is divided into two portions. The diffusion occurs both to the The diffusion to the right hand side is considered first in the following.t 0 PA £3 (Current) 2 1 constant ) where 1 = (Gradient) = U3 Lx/cose. p4A and (lp)4A and appears at the points Al and A 2 .
Geometry for explanation of the crossflow diffusion in a rotated two dimensional Cartesian coordinate . Ya.68 ti .4. 4..ct t2 t3 Fig.
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. D D = uAx(lp) = VAy p (4.39 is identical to the following simple expressions. sine cose RHS LHS cos1 sin(+8) (4.38 is the effective diffusion constant when the diffusion is confined to the one dimensional fluid element.38) Equation 4.40) . There are two diffusion components in a two dimensional problem as follows.39) Equation 4.69 Repeating the same procedure for the left hand side we can find that the right and left hand side diffusion constants are equal to the following. 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. The crossflow diffusion constants are the ratios of the current and gradient of 4 in the x and y directions.
38 should be differentiated with respect to 6 in order to find out the velocity direction with maximum crossflow diffusion. D DDM = UAlxy sin2e= 3 3 4(by sin36 + Ax cos36) UAx tan6 1 sin26 3 4(tan8 sin6 3 + cos3) (4.70 where p = u Ax u v By Equation 4.41) Equation 4.42 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. < 6 < 81 or < .42. 4. 4. 4.42) Equation 4.6.6 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. In Fig. When both 6 and 61 are equal to .41 shows that DRH S (or DLH S ) is maximum at an Tr7LHS angle of 61 < 6 e between 4 and e1. In the donor cell p' y scheme the value 4 is divided into three portions px 0 .5.40 and Eq. 4.42 underestimates the crossflow diffusion as shown in Fig. (4. that is to say. Eq. 4. 4. 3 0 when tan 6 = tan6 1 .
0 150 300 450 600 750 900 Fig. 0.5 80 \ % % 4 1.0 0 0 .71 Cross flow Diffusion 2.5 I I 0. Comparison of the prediction formulas for crossflow diffusion constant by De Vahl Davis and Mallinson and Huh .5.0t Dcf UAx de Vahl Davis & Mallinson Huh 81=8C 1. 4.
72 z C V Az X 0 o a X AY Ay B y x z C p . 4. p Fig. Geometry for explanation of the crossflow diffusion in a three dimensional Cartesian coordinate .6.
£1 = AX =/(x4cosa)2 + (4cos)2 4cos) 2 = Ax 2 + 42 2Ax£ 4 cosa . S= 4 + y z zAx 1 cos+ + cos Ay Ax + cosy Az (4. 4. B. BX. cosa cosB cosy (4. XC. and C instead of at the point X. n C denote The coordinate of the Any point on the plane ABC should satisfy the following Eq.73 z occurring at the points A.43. menon. CX which are denoted as £i' z2' k3 can be obtained as follows. XB. S+ Ly Ax + Az = 1 plane ABC (4.and nA unit vectors along those directions.43) Since the point X is on the plane ABC. t4 cos8. X(£ 4cosa. £4 cosy) nB .45) The lengths of AX. the following relation holds. along the directions XA.44) 4 Therefore. This is the socalled crossflow diffusion phenoThe currents due to the crossflow diffusion are . point X is given as follows.
47) etc.OX (4. n C can be obtained in terms of their components in x.48) The currents in the directions nA . nB. Therefore.46) + The unit vectors nA . z directions. I (Current)xAi XAI = px Px C1 1 Lt . nB' n C can simply be given as the products of the transported quantities and velocities. y. (Ay4 cosS). 2 o C 14cosy] = 1  £ 4 cosa. 4 cosy] 1 nB = [i4cosa. Z 4 cos8.  n A = XA +1A  OA .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. (Az 4 cosy)] (4.
(4 pz x co s a ) ct 4 + at ( 4 cosa) 4 .51) p Ax D x = lt (Axk 4 cosa) + p Ax .50) Now the x direction crossflow diffusion constant may be given as the ratio of the current to the gradient of 4 in the x direction. z directions are given as follows.75 I(Current) XB = Py p2 t I(Current) XCI where At = k /U z '3 At (4. (4. Px D (Ax£ + Py D + nC(4 (Current)x 4 cosa) c osa) At ( 4 (cosa) (Grad)x = 4/Ax Therefore./ Ax (Grad) = 4.49) The gradients of 4 in the x. y. (Grad)x = 4./ Ay 4/6x Q/by (Grad)z = 4/Azl (4.
Dy = vLy(lpy) Dz = wAz(lpz ) The rotation of (4. p Ax D = ux[ £4cos x .54) 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. etc.52) Inserting the expression for At given by At = £4 /U.53) The crossflow diffusion constants in the y and z directions can be obtained in the same way. .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. finite element method. tensor viscosity method. This section is a review of the schemes that The have been used to eliminate the numerical diffusion. although none of them has been successful enough to be accepted widely.1 = uAx(lpx) (4. most important ones may be the skew differencing. indices will also give the same result. we can simplify the result as follows.
The donor cell treatment of the convection term results in a 5point relationship in a two dimensional problem... the interpolation between neighboring points may still give some crossflow diffusion. Since the upstream point in the velocity direc tion does not necessarily fall on the mesh point where the quantity under consideration is defined. Basically the inclusion of four additional corner points contributes tomore accurate numerical modelling of the convection.^ILli.~_ 1 i_.4 fall in the category of a corrective scheme.1 can be categorized into two types. but just appreciably lower than that of the donor cell scheme. In the skew differencing scheme the 5point relationship becomes a 9point relationship..3. but directly along the velocity direction. 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.II1~LI1I~~ ICII1_1___~1___*_____11__ 77 Most of the schemes in Table 4. In the skew differencing type schemes finite differencing of the convection term is done not in terms of the x and y directions. However.. The crossflow diffusion is not eliminated entirely in the skew differencing scheme. S. skew differencing and corrective schemes.. The corrective scheme is to reduce the diffusion constants in order to compensate for the additional false . the interpolation should be done between neighboring points to obtain the value at the upstream point.
inte Conribo:r/me.78 Chss' iction Cen:ered .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 !. Esposito Glas and Rodi Huffenus Lnd Khali./LLUE Huint/QUICK Wilkes/LUE Lillimpon/SUD LiiL.thod of chaaceriics Fiite anuryic method Tenso. £Ebzhau t aL/PATANKAR Lilinpion Orlandi Priddin Wada cit l.Highe cdtr upwi:nd :"'ie Cdi't.j. upwind f~iite di Teretn atl.e elements Clifft Dontcl t al. 1981 [65) .ronLUD Lillinion/VUDCC Nuana.hod Mesh *txLnsformation rr.thod LLxW'en doff  21 x11 80 x 400 Table 4.1.ad.1QUICK t ETovh ci . List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research.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 . Wilkes Elbahu ct l. Grandotto Hickmott Firs.ordc. viscosity method Sel.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 ..aptjve me.
53 and Eq.53 and Eq. although the validity of its correction formula is not clear. where the balance equation is set up for a given control volume by considering convection and diffusion at the control surface. y and z directions will be reduced by the amounts of D . stants in x. and D z in Eq. 4.1 Raithby's Scheme Raithby's scheme [53] is an Eulerian type of approach.54. The tensor viscosity method is one example of the corrective scheme. Since the crossflow diffusion is dominant over the truncation error diffusion and the effective diffusion constants for the crossflow diffusion can be predicted theoretically. D. It is therefore a conservative scheme. 4.54. 4. 4. the corrective scheme can give a numerical solution which is almost free from numerical diffusion. The correction formula The diffusion con used here is Eq. The corrective scheme is valid only if the effective numerical diffusion constant can be predicted accurately. 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..79 diffusion. 4. 4 = C1 + C2 n = C1 + C2 (  (4.55) where n is the distance normal to the flow and U = /u 2 +v 2 .3.
the calculation domain showing grid lines. 4.7. control volume and notations in Raithby's scheme . A segment of.80 Fig.
81 The convection quantity at the control surface is calculated from the upstream value in the flow direction and the upstream value is obtained by interpolating two neighboring points.56) where U = /u is the coordinate in the velocity direction and 2 +v 2 . The inclusion of the four additional points also complicates the structure of the coefficient matrix in an implicit scheme.57) where the upstream value €US may be calculated by .3.= U ax dy SC (4. u 4 + 'v. 4. which may be too complicated and time consuming for a practical purpose. Although this scheme is nonconservative.2 Skew Differencing By Huh Skew differencing scheme here treats the convection term in a Lagrangian way. The convection term is differ enced in the velocity direction between mesh points. not in terms of the convection that occurs at mesh interfaces. This should be repeated for the four control surfaces of every mesh in a two dimensional problem. 'ij US (4. it is much simpler than Raithby's scheme and becomes conservative in a unidirectional flow.
3.4 Corrective Scheme By Huh The corrective scheme is to use the reduced diffusion constants. 4. DDx. D are derived . 4. DD y .3. The validity of this scheme is based on the fact that the crossflow diffusion can be predicted accurately.3.58) t2U= Eq.54. Dx = uAx(lpx) D Yy = Ay (lpy ) The expressions for . uv (4. z directions to compenx sate for the crossflow diffusion effect.2. the crossflow diffusion constants D in section 4. D . y.2.82 interpolating two neighboring points. the tensor viscosity [25) T = Lt uu u2 The expression for is given in the following. 4. DD z in x. Some of the results of this skew differencing scheme is given in section 5.53 and Eq. 4.3 Tensor Viscosity Method The tensor viscosity method treats the numerical dif fusion constant as a tensor quantity.58 for the numerical diffusion constant is apparently not consistent with the results in Eq. 4.
These two become identical when the flow field is uniform or a very fine mesh is used so that the flow change over neighboring meshes is negligible. which is used with the cell dimensions to get the crossflow diffusion constants. 4.+ . In a coarse mesh or recirculating flow field the latter implementation strategy is definitely preferred. 83 D z = wAz(lp z ) where px = u (4. they should be averaged again between neighboring mesh points to calculate the crossflow diffusion current at mesh interfaces.1 Mesh Point Implementation The mesh point implementation is to assign the crossflow diffusion constants Dx .Dz at the center of a mesh where all quantities are defined except the flow field.4. . After the crossflow diffusion constants are assigned to every mesh point.+ w Ay Lz x etc.59) v u ..3. The velocities at two boundary faces are averaged to get the representative velocity._IIU__lpULI~~I___~lLlt i~Lli ~i ~iL~pn~sl. Dy . Two implementation strategies have been tested so far. the mesh point implementation and mesh interface implementation.
Vl) and (U2. For the first two cases the cross flow diffusion constants are zero because the velocity vector is aligned with the mesh configuration. and the resulting D x and D y will be assigned to each interface. The right interface will have the crossflow diffusion constant (Dx1+Dx2 ) and the top and bottom interfaces. so that they are proportional to V1 and V2.2 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. Dyl and Dy 2 . U and V. The two velocity sets (Ul. The third case is expected to have a flow split as shown in the figure.84 4. There are three possible cases that should be treated independently in a two dimensional case. For the second case the two outward velocities.V2) are used to give (DxlDyl) and(Dx 2 'Dy2 ). can be directly used with cell dimensions in Eq. 4.8.59. as shown in Fig. U1 and U2. 4.3. Note that the crossflow diffusion constant for the interface with an inward velocity does not have to be considered in that mesh because it will be considered in the neighboring mesh. The mesh interface implementation is always recommended although it has a more complex program logic than the mesh point implementation.4. The velocity U is split into two components. The same logic can easily Appendix C includes be extended to a three dimensional case. .
Three possible flow configurations for mesh interface implementation of the corrective scheme in a two dimensional problem . 4._~I~~ _~~~~_I__~__~~ __ 85 (1) (2) (3) Ul U2 V2 VI U U V14V2 VI+V2 V2 U2 = UV2 U = Ul + U2 TV2 Fig.8.
86 the fortran program for calculating crossflow diffusion constants in an arbitrary flow field. 3.60) Explicit (lD): n+l I n i At + u (i i x n i. Implicit and ADP schemes There are three typical solution schemes. 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. Eq. Explicit.) = i1 n D Ax ( i+1 2i l n n + n ) + iS i1 (4. 4. 4.60. conservation physical constraint accuracy stability 2 ax 2 2 y 2 +ux +at + 7 + T7) + S sz (4.4 Comparison of Explicit. The pros and cons of the three schemes will be compared in the following four viewpoints. 1.62) . which is a parabolic partial differential equation. 4. for the general conservation equation. 2.
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. or affected only by the source term contribution (SO). when the net inflow or outflow at .87 ADI (2D): n+ ij (A) _n ij In+ + u ij _ cn+ in + v n ij1 At 2 Ax Ay )n+ i+lj 24 n+ Ax 2 +n+ n l + D j+l n ij 2 Ay n+l ij n iji + S n+l ij n+ ij n+ ) ij n+ i1 n+l ij1 At 2 Ax Ay n+ =D i+i S2..4. 2 Ax n+l 2 * ij v n+l ) ij n+l Ay 13j1 n+1 ij+l n+l 1j1 At Ay2 + S (4.1 Conservation The conservation principle is that the total amount of 4 should be conserved (S=0).63) 4. + 4).
the same exchange term should be used for neighboring meshes. where the convection terms of momentum equations are linearized in a nonconservative way. A scheme may be both stable and nonconservative and also both unstable and conservative because the conservative property is independent of the stability. Although they usually give .4. the conservative property may not be crucial.88 the domain boundary is equal to zero. Although a conservative form is preferred to a nonconservative one.2 Physical Constraint The physical constraint is closely related with the stability in a practical sense. All the explicit. When we are concerned with a steady state solution. In order to guarantee conservation. the conservative property is as important as the stability because the error from nonconservation may accumulate over many time steps. however. implicit and ADI schemes can be made conservative by using appropriate differencing forms for the convection term. For a long time transient. + CI 4. there are some nonconservative solution schemes like ICE (Implicit Continuous Eulerian).
89 similar constraints on the time step size and mesh spacing.1 Diffusion There is a maximum net flow that can occur between neighboring meshes by diffusion. The laws of physics impose some restrictions on the numerical scheme to get physically reasonable solutions. R C AxAy = CAxy + D R Ax Ax CAy C y At At ¢4AxAy = 1RAxAy .4. 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. they are from entirely different origins.D . new values of 4 are given as follows.2. 4.
65 is used. Therefore the physical constraint should be respected in addition to the stability condition to get a meaningful transient solution.64) Eq. 4.65) Ax Equation 4. however.64 may be extended to a ddimensional case as follows.65 may be a too conservative criterion in a real problem. the solution will be unstable or stable with damping oscillations. 4. 4. there will be a damping oscillation in the solution when the time step size is greater than that given by Eq. we have the following physical constraint. R AxAy(4) C) > 2D (R Ax CyAt 2 DAt X2 < 1  (4.64 or 4.64 or 4.90 Since the relationship.65. d 2D t < 1 (4. scheme given in Eq. should still hold. 4. When the physical constraint is ignored and the time step size greater than that given by Eq.63 is unconditionally stable. t' > Q' . Both of the oscillation and instaThe ADI bility should be avoided in a transient problem. .
than the Courant limit.2.3 Accuracy Realistic interpretation of a finite difference equa tion comes from an integral form of the conservation .B 4y S The explicit and donor cell treatment of the convection term gives the convection quantity at interface S as ulAAtAy.91 4. Then uAttay is no more an appropriate expression for the convection quantity during At.4. the damping oscillation or instability may occur as in the case of diffusion. a portion of the control volume C will cross the interface S.2 Convection C A u B C A ).4. It is based on the assumption that 4A is the repreIf At is greater sentative value of the control volume A. At < 'x u (4. Therefore.66) When a time step size greater than the Courant limit is used. 4.
I I t i r _ I I [ n n+1 n n+l _ 0 It Explicit Implicit CrankNicholson $i+. 4.92 n+l n+l 4' I I.9. ~~CIIL Is CIIII r r I I I I I! US Ds > I r I US DS Donor Cell Central Differencing Reverse of Donor Cell where US(Upstream) DS (Downstream) Fig. Profile assumptions of Q in space and time for various differencing schemes .
whether the error introduced at a certain step will increase or decrease as the calculation goes over to the following steps. It depends on the maximum eigenvalue of the iteration matrix.4 Stability The stability is a mathematical problem. The Von Neumann . appropriate schemes should be chosen according to those numbers or weighting should be done between neighboring time steps and mesh points to increase the accuracy of the solution. the accuracy depends on the profiles 4 in time and space because the exact profiles assumed for are not known a priori. The infinitesimal approximation of differential terms is not appropriate in most problems of our consideration.93 equation. accuracy depends on how well we can compute the convection and diffusion quantities at control surfaces over the time At. In other words.4. The accuracy expressed in terms of O(At). Since the exact profile depends on the Peclet and Strouhal numbers of the governing equation. 0(Ax 2 ) is The also not appropriate for finitely large At and Ax. However. The stability can be checked by the theorems about matrix eigenvalues or Von Neumann analysis. the accuracy is usually the least important among the given four considerations. If its absolute magnitude is less than one. 4. the error propagation is suppressed and the largest error contribution will be from the previous time step.
68) .4.2 2i + D i+1 i n 2 i i1 (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. but its algebra may get complicated. (4. 4.4.1. n+l i n €i+i I + u i+1 n ii 2Ax i_1 lt n n i+l . stability criteria will be derived for the explicit. The stability is not a sufficient conIn the fol dition for an acceptable numerical scheme.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. lowing sections.67) The solution 4.. 1 4n = E~nei where I = . 4. implicit and ADI schemes using the Von Neumann analysis. is expanded in Fourier series as follows.4.1 Explicit Scheme The Von Neumann stability analysis of a finite difference equation is simple in its idea.
in Eq. the function . 4. 4. f(t) = i 12 = (4d2_ 2 )t2 + (Sd 2+4d )t + C 2+4d 24d (4.68 is substituted results. [i .72) In order to get the stability of Eq. 2 11 = [1 . 4. mode to achieve stability.71) The function f(t) will be defined as follows.67.95 The absolute value of should be less than one for every When Eq.70 may be rewritten as follows by defining cos6 x as t.70) Equation 4. the following expression for = lCxsin xI where cx d x uAt A tx  2d (lcos x) (4.2dx(t) where t = cose 1 < t < 1 ] 2 +C 2(1t 2) 1 (4.2dx(cosex) 2 + [Cxsinex 2 < 1 (4.67. Therefore.69) Dbt Dt 2 Tx.
f(t) < 0 (4.73) for any t in [1.74) One of the two meeting points with the taxis is the point t=l because f(l) is equal to zero. 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.75) The parabola f(t) may have three distinct shapes according to the sign of the coefficient of the second order term.2d )2 > 0 (4.10(1).1]. Therefore. The parabola f(t) always meets the taxis because the discriminant is always greater than or equal to zero.73 can be solved by a graphical method using the characteristics of a quadratic equation. .96 f(t) should be less than or equal to zero for any t in [1. D(Discriminant) = (C 2 . f(1) = 0 (4. 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] Equation 4.
3 0 0 /1 i2 2 4d 2d x x 22 4d c xx t / (1) 4d 2 c 2 >0 x x (2) 4d 2 c =0 x x 2 2 (3) 2 2 4d2c2<0 x x Fig.97 f(t) f (t) f (t) 'I I 4. 4. r I I t i t 1 1 t 0. Three distinct shapes of the parabola. f(t)..10.. according to the sign of the coefficient of second order term .
C (4.dx) is shown in Fig. 4.4.98 f(l) = 8dx(2dx1) < 0 0 < d < (4. 4.1. 4d 2 . 4.2d Axis of symmetry: t = 4dx dx > C 2 x > . 4.76 and Eq. .10(3).4.11. t=l as shown in Fig.77.2 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.78 is a finite difference equation with donor cell differencing of the convection term. In case (3) the parabola f(t) is concave downward and the axis of symmetry should be on the right hand side of the point. we get the following stability conditions.76) In case (2) f(t) is a straight line and f(l) again should be less than or equal to zero. 4.77) Summing up the results in Eq. 4.
99 d x 1 1 0 1 dx = C2 x 2x C Fig. Stability ondition in the plane (Cx.11. 4.d ) for an explicit scheme with central differencing of convection term in a one dimensional problem .
100
n+l
i
n At
At i
n
+ u i Ax
n
ii
n
= D
i+l
n
i
n
ii
Ax
2
2
(4.78)
where the velocity u is positive. The Von Neumann analysis of Eq. 4.78 gives the following stability condition.
2
It
1(C x +2d [
x
)(1cose x ) 2 + [C sine ) x
1
(4.79) (479)
can be shown that the value of IC12 in Eq. 4.79 does not
change when the velocity u is less than zero, if the Courant number, Cx, is defined such that it is always positive. The function f(t) is defined as follows to make it easier to solve Eq. 4.79.
f(t)
=
12
1
S[(C x+2d)
2Cx
]2 t
+ [2(C x+2d )2 2(C +2d
) ]
+ 2(C x+2d )]t (4.80)
+ [(C +2d ) +C
The function f(t) in Eq. 4.80 should be less that or equal to zero for any t in [1,1] for the stability of Eq. 4.78. f(t) < 0 for any t in [1,1] (4.81)
The parabola f(t) meets the taxis at the point, t=l. f(1) = .0 (4.82)
101
Equation 4.81 canbe solved by the graphical method as in the previous section. 2 2
(C +2d )
 C
> 0
:
0 < Cx +2d X < 1 
(4.83)
(c x +2d x)
2
C
2
x
< 0
d
>(C 2C ) dx x
(4.84)
The results, Eq. 4.83 and Eq. 4.84, are shown graphically in Fig. 4.12.
4.4.4.1.3
Two and Three Dimensional Donor Cell Differencing of Convection Term
The two dimensional extension of Eq. 4.78 is given in the following. n+l ij At n D . i+D 3 24
Lx
n ii
n + u ij Ax n
n i1 n
n + vii by
n iji n
+ 4
A2
1il3
n n 4  24 + D2+l ,_2
ay
+
ii
(4.85)
The Von Neumann analysis is applied to the finite difference equation, Eq. 4.85, yielding the following result.
 1 = C x (le
lex)

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

Ax
C
 vat
d

DAt2 x2
x
Y
DAt Ay2
The Courant numbers, Cx and C y , are defined so that they are always positive, independent of the directions of the velocities u and v. Equation 4.86 can be simplified
by defining al and a 2 as given in the following Eq. 4.87. = 1 + al1 where a= 1 C x  2d x + (Cx+dx)e x I6
x +
a2
(4.87)
+ de x
le
x
S
=
2
C
Y
2d
+ (C +d )e YY Y
I6
x
+ de Y
Iy
The stability condition is that the absolute magnitude of 5 should be less than or equal to one for any values of 8x
e and By.
From Fig. 4.13 and Fig. 4.14 it can be seen that
indea1 and a2 should satisfy the following conditions pendently. a + 1
4
(4.88)
(4.89)
104 Im \ . . 4. The region in the complex plane where (a +c ) should exist for the stability of an explicit conscheme with donor cell differencing of vection term in a general two dimensional problem .1 SRe 2 al +2 1+a +a 2 Fig.13.
Fig. .. The region in the complex plane where the amplificatior factor.14. .105 .Re Both al and a 2 should be in the shaded circle to ensure that Ial+a 2 +1I< i. 4. should exist for the stability of an explicit scheme with donor convection term in cell differencing of a general two dimensional problem .
4. Cx and d . and time step. A three dimensional case can also be solved through the same procedures and the results are given in Fig.1.88 and Eq. which depend on both the mesh spacing.90) a 2 = C [1 y R R Y + R (4.e Rx . 4. The stable region shrinks on a linear scale with dimensionality.4 Stability Condition in Terms of Cell Reynolds Number The stability conditions in the previous three sections are in terms of the dimensionless numbers.15. At.4. the stable range of Cx should be determined in terms of the cell Reynolds number Cx/dx'.89 and al and a2 may be rewritten as follows. our usual concern is can be checked directly. When Ax and At are already given. Therefore. the stability However. 4. Ax.91) . The stability condition is again given by Eq. At.106 Both Eq.16. which is independent of the time step size.d) should be in the shaded region for stability. I 1 2 + (1+)e Rx Rx 2 + (1+)e a1 = Cx [1 + Ie 1 3 . where the points (Cx. 4. the stable time step size for a given mesh spacing.e Y) (4.dx) and (C y. 4.4. 4.88 and 4. The results are given in Fig.89 can be solved by the graphical method as in the previous sections.
dx) and (Cy. . Stability cc.15.107 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. 4.idition in the planes (Cx.
4.dx) etc.16.108 d xy(d or d z ) Cx d dx + 1 1 Cx (C d! S3 2 . for an explicit scheme with donor cell differconvection term in a general three encing of dimensional problem .(C Cx x y or Cz) z d ) Fig. Stability condition in the planes (Cx.
90. [ .94 ) D(Discriminant) = C 2 (K2C ) 2 > 0 (4. (1) Cx 2 K2 1) > 0 concave upward .C K(lt) + C ( l_1t = (C 22 KC 2 2 )t2 + (2C 2 22 22 K +C K)t + C 22 K  C K + C 2 < 0 (4. 4.93) ) where 2 K = 1 +. 4.95. f(t) = C (1+K) 2 (lt) .92) Equation 4.) RX (1cose )) x + [C sine] x x < (4. and has distinct shapes for the Following three cases.94 and Eq. 4.C (+2x . 4.R and x t = cosex The stability condition is now that f(t) should be less than or equal to zero for any t in [1.95 ) The function f(t) satisfies Eq.1] f(l) = 0 (4.88 and Eq.__LI___IIII__LI*X__ ID I 109 The following condition for Cx and Rx can be derived from Eq.92 can be simplified as follows.
therefore the subscript x may be replaced with the subscript y in Fig. 4. solution procedure is similar to those in the previous sections and the results are given in the following as. to zero.17.110 (2) C 2(K 21) = 0 linear (3) 22 C 2(K 2_1) < 0 concave downward For case (1) and (2). although it can be relaxed if there is a constraint on the velocity direction and mesh spacings such that the following' relation is satisfied. The same result will be obtained for the ydirection.96) R 1 x < 0 < C x < (1+2) Rx . R > 0 x : 0 < C < x 2 2 2(1+) x and (4.97 are shown in Fig.17 is applicable to all general 2D problems. t=l. 4. 4. The stability condition in Fig. 4. V f .17.98) Ax . 4. (4.yu < 1 (4. 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.97) The results in Eq.96 and Eq.
Cy) for an explicit scheme with donor cell ditferencing of convection term in a general two dimensional problem . 4.Cx) and (Ry.111 x (C y C x  2 1 2 (1+ L) R x C = 2(1+ ) x 2 0 R (R x y Fig.17. Stability condition in the plane (Rx.
18.Cx) for an explicit scheme with donor cell differencing of convection term in a two dimensional problem with f v u/Lx . Stability condition in the plane (Rx. 4.112 1 1 x CX x 1 (l+f) (1+ x ) 1+f 1 1 f 0 x V where f =y u Ax Fig.
uLx.17 and numerical experiments. 4. 4.Cx d > . parison between the analytical result in Fig.2 shows the com 1 l+f in Fig.19 and Fig. 4. the stability condition is found to be the following.100) For a general two dimensional problem. 4. 4.17 should be Table 4. (4. d >  greater Cx (4.4.113 Then the number replaced with on the C axis in Fig. 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.C Y y .20.102) dx > . 4.18.4.99) (4. and it can be seen that the implicit scheme is unconditionally stable if the diffusion constant is than .101) (4.99) The stability condition is shown in Fig.2 Implicit Scheme The following Eq. 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.
380 0.060 Table 4.333 4.2.333 833.833 8. R x 0.333 83.120 0.019 Numerical Experiments 0.403 0.488 0.167 2.499 0.780 0.083 Von Neumann Analysi s 0.260 0.3 8.114 Maximum Stable C.720 0.660 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 .120 0.147 0.180 0.540 0.660 0.629 2.
C x 2x x 2x C x dx 1(C2+C 2xx ) Fig. Stability cndition in the plane (Cx.115 d =  x d =.19.dx) for an implicit scheme with donor cell differencing of convection term in a one dimensional problem . 4.
20. 4.116 Cx=1  2 Rx 2 1 0 R x R C x R +2 Fig. Stability condition in the plane (Rx.Cx) for an implicit scheme with donor cell differencing convection term in a one dimensional of problem .
4. 2 _C C 2 {[(cos )(+d X))] C [sine I2} {(1cos6 C 2 )(.4.63 gives the following stability condition.117 4.+d )] C ) {[1+(lcosex +dx)) +[fsinex }{[l+(1cosy) (+dy) + Y[siney } (4.4.3 ADI Scheme There are two possible formulations of the Alternate Direction Implicit (ADI) scheme given in Eq.63. 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.104) . the following condition should be satisfied. The Von Neumann analysis of the first formulation in Eq. C d x>2 (4.103) jsj <1 In order for I. 4.jto be less than or equal to one. One of the formulations has both the x and y direction components in each step.
.63 can also be given by the Von Neumann analysis as follows.63 has the same stability condition as the fully implicit scheme. 4.106) C d y  >  ~ 2 (4.118 The same condition should hold for the ydirection.107) Therefore the second formulation of the ADI scheme in Eq. C d x > 2 2x (4. 4.105) The stability condition of the second formulation in Eq. C d y>  2  4 (4.
1. 5. Natural convection occurs due to the heat transfer between the obstacle and containment air. the ADI scheme is explored to use a time step size larger than the Courant limit to save the overall computational time. density is determined by temperaSince the natural convection of air ture and pressure. 5. occurs as a result of the buoyancy force due to density .2.1 Natural Convection Results The containment is modelled as a two dimensional rectangular compartment with an obstacle as a heat sink. 5. Updating of the reference state. energy convection and heat transfer models are considered and the resulting flow fields are given in Fig. For example.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. 5.1 and Fig.119 CHAPTER 5 RESULTS Some calculations are performed to validate the physical models and numerical schemes in the previous chapters. Finally.
5.5 ft/sec h = 1.8 sec . 0.1. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 10.4 sec AT = 100F T .0 Btu/lbm 0F (Q ob = 208. = 680F air T = 580F Fig. = 677.120 p ).29 Ibm A= 2 ft2 T.8537x10tu/ft20 F sec (Cp) = 1.
5 sec .0 Btu/lbm OF (c )b = 208.29 ibm (K 2 A= 2 ft T.2. 5 ft/sec h = 1. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 62. T ab = 580 F Fig.4 sec 1 0 AT = 10 F T air = 680F .8537x10 tu/ft2 F sec p)ob = 1. = 677.121 > 0. 5.
1) T Pnew = old + Ttotal c v M f. however. new Pnew RT new (5. time Pressure of the fluid at old and new steps. depends primarily on the temperature. not on the pressure because of a high modulus of elasticity. Told' Tnew: Temperature of the fluid at old and new time steps.cell where Pold' Pnew: Density of the fluid at old and new time steps. The equation of state of an ideal gas is given in the following. c : Specific heat of the fluid at constant volume. Pold' Pnew: .122 gradients.total Pold Told new old Qcell cv Mf. temperature and pressure are equally important parameters in determining the flow field. The density change of liquid.
cell..total : Mass of the fluid in one cell and in the total domain. . but they may also be based on the vertical height of the obstacle. Qcell total : Heat input to the fluid in one cell and in the total domain. 5. 5. The Nussett and Grashof numbers here are based on the axial length of one computational mesh.4 for air. .0 Monatomic. The value of k is 1.2 in terms of the dimensionless numbers. 5.4 when internal energy is used for enthalpy in air flow. k = c /C for liquid and 1.2 Energy Convection Another difference between liquid and gas is the ratio of the Specific heats.1. diatomic and polyTherefore. An experimental correlation is given in Eq.3 Heat Transfer Modelling The heat transfer to the obstacle is modelled as a natural convection from a vertical surface. convection energy transfer should be in terms of enthalpy instead of internal energy. The energy convection will be underestimated by a factor of 1. Mf. The VARR calculations 121] have shown that the magnitude of velocity can be affected by a factor two by updating the reference pressure at every time step. the atomic gases have different values of k.1.123 Mf.
25 0.8537 x 10 .876 The heat transfer coefficient is given as. This may be a typical value for the natural convection heat transfer in the containment without any phase change.2179 x 10 B = 1.2) 3 2 Gr = gS(AT)L p 2 (5. L = 2 ft AT = 500 F p = 0.124 Nu = C(GrPr) n ( 5. .021 n 0. Gr = 9.0763 ibm/ft 3 g = 32.555 0.3) Pr = 0.sec .923 x 103 Therefore.7134 x 108 Nu = 89.2 ft/sec' 5 V = 1.4 The following data are used to calculate the heat transfer coefficient at the obstacle wall.708 (air) GrPr 105 .109 > 109 C 0. h = 1.4 lbm/ftsec Btu/ft 2 F.
Heat transfer area between the obstacle and fluid over one computational mesh. pf Pob: c h : Fluid density. can be obtained from heat balance between the obstacle and fluid as follows.2. 5. Obstacle density. T = 677. 5.1 and Fig.1.4 Turbulence Modelling Laminar flow is assumed in obtaining the results in Fig. T. Specific heat at constant pressure. the heat . CQ = Cpob (TT) (5. If the flow is turbulent. Heat transfer coefficient between the fluid and obstacle. Therefore.125 The VARR input [21) requires the time constant of the heat transfer instead of the heat transfer coefficient.4 sec. The time constant. (Mc )ob (MC) obf hApo b ob f (5.5) where M A : Mass of one computational mesh.4) where CQ is the source term in the energy equation of fluid. 5.
5. 5. 5.2. is compared with De Vahl Davis and Mallinson's in predicting the magnitude of crossflow diffusion. Figure 5.3. The flow in Fig. The skew differencing and corrective Huh's formula schemes are tested for various problems.126 transfer.7 show that excessive numerical diffusion has occurred due to the .5 show that the numerical solutions is close to the analytical solution and the truncation error can be neglected. 5.1 Truncation Error and Crossflow Diffusion The truncation error diffusion is compared with the crossflow diffusion in a two dimensional.3(A) is parallel with the mesh orientation so that no crossflow diffusion occurs. The results in Fig.6 and Fig. The flow in Fig. steadystate problem in Fig. 5. 5. The corrective scheme is tested in recirculating flow problems for mesh point and mesh interface implementations. 5.3(B) is skewed at 450 to the mesh orientation and both truncation error and crossflow diffusions are included in the solution error.2 Numerical Diffusion The truncation and crossflow diffusion errors are compared in a simple geometry where an analytical solution can be obtained.to the obstacle will be enhanced and the diffusion transport of momentum and energy will also be enhanced in the containment air. 5.4 and Fig.
3.0 a= 1. 0 ST By =0.0 B T specified Fig.127 (A) .0 u = 10. (A) parallel flow and (B) cross flow.0 ay AB=BC=5. Problem geometry for the two cases.0 u = v = 7.0 BT S=0.0. for evaluation of numerical diffusion .0 x BT  @x a( aT + 2 2 a2 ay ) Th= 10. 5.0 B T_ 0.07 (B) =0.0 (A) (B) AB=BC=5.
4. Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 10x10 meshes along the line CD in Fig.128 Analytic solution Numerical solution with Donor Cell T 10. spatial coordinate Fig. 5. 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.__~ .~ 129 Analytic solution Numerical solution with Donor Cell C spatial coordinate Fig.^XI ill I~LX_L~LI.__LL l. 5.~l r.~ I~.3(A) . 5.
130 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Df cf T 10 spatial A coordinate Fig.3(B) . analytic solution with increased diffusion constant and numerical solution with donor cell differencing convection term in 10x10 meshes along the of line CA in Fig. 5.6. 5. 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.131 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Dcf T 10 / I / / . 5.3(B) s' / line CA in .J 0 I' C spatial A coordinate Fig. Comparison cT the analytic solution.3(B) . 1/ 5.. . 5./ Fig.7.
5.8. 5. 6 = 6 = 450.6 are for 5. Raithby's and Huh's. 5. 5.53 and Eq. It can be seen that the former result is much better than the latter. 5. 5.5 and Fig. Figure 5.2.6.11 and Fig. 5.4 and Fig.9 is the result by Huh's formula and Fig. solution. The problem geometry in Fig.132 crossflow diffusion. 5. 5.7 than in Fig.2. De Vahl Davis and Mallinson's and Huh's formulas give the same crossflow diffusion constants for the case.6 and Fig.2 Validation of Huh's Correction Formula The crossflow diffusion can be predicted by Huh's formula. 10 x 10 meshes and Fig.7 show that the numerical solution can be reproduced by the analytical solution with increased diffusion constants by the amount given by Huh's formula. Figure 5. it is necessary to test them for various cases. Since the corrective scheme is based on the validity of those correction formulas. more diffusion has occurred in Fig.3 Skew Differencing Scheme Two skew differencing schemes.10 by De Vahl Davis and Mallinson's for the problem in Fig. Figure 5. 4. 5.8 has the arbitrary angles of 6 and 61 such that 6 = 600 and 61 76. 5.810. 4. Eq.7 are for 6 x 6 It is seen that a finer mesh gives a better Since the crossflow diffusion constant is pro portional to the mesh spacing.12 for a simple pure . are tested in Fig. 5.54. meshes.
0 Ax = 0.0267 u = 5.2406 by = 1. 5.0 v = 8.7674 DH u h = 3.335 Huh A e1 = 76.8.810 8 = 600 8 Meshes 24 Meshes Fig.133 Yo 2 T=0 D a = 1. 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. 5.134 Analytic solution with D+Dcf Numerical solution with Donor Cell T 10.9. 5 0 spatial coordinate Fig.8 . Comparision of the numerical solution with donor convection term and cell differencing of analytic solution with increased diffusion constant. by Huh's formula along the line BD in Fig. (D+Dcf).
by De Vahl Davis and Mallinson's formula along the line BD in Fig. Comparision of the numerical solution donor cell convection term and analytic differencing of with increased diffusion constant.135 Analytic solution with D+DDM Numerical solution with Donor Cell T 101 spatial coordinate Fig.10. solution (D+DDM). 5.8 . 5.
136 Donor cell 100 100 100 100 100 93. Comparison of the true solution and solutions by donor cell scheme. 5.25 65.75 81.5 68.63 87. 0 100 100 100 100 100 50 100 50 0 50 0 0 V Fig.75 25 U 12.25 18.11.25 0 Skew differencing by Huh (True solution) 100 100 100 100 50 0 0 0.5 0 6.75 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. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with e = 45* .
85 0.6 0 Skew differencing by Huh 100 90.11 0 11.47 23. 5.3' 40.9 33.6.5E 25.74 38.25 0 0 0 . 62. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure covection problem with e = 63.13 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 6.430 .26 80.01 55.22 9.71 20.25 1.E 21 0 12 True solution 3.59.5C 100 100 50 Fig.1 6.137 Donor cell 100 100 100 100 50 .2! 90.25 53.91 71.88 81.33 70.70 4.99 10.50 2.A Skew differencing by Raithby 100 100 100 100 96.3E 21.67 5.5 0 0 63.83 72.3: 11.43 7.11 3.8922.96 17.52 1.12.91 31.430 0 50 0 Comparison of the true solution and solutions by donor cell scheme.86 16.25 37.56 1.
the However. The purpose of the corrective scheme is. Raithby's scheme is an Eulerian Therefore. 5.3.430 angle and neither Raithby's nor Huh's scheme can reproduce the true solution.4 Corrective Scheme The prediction formulas of the crossflow diffusion constants are validated by showing that the numerical solution can be reproduced by the analytical solution with the appropriately increased diffusion constant.3. 8 = 1e= 45*.13 and Fig.12 has a flow at a 63. method while Huh's scheme is Lagrangian. Figure 5. to obtain the true solution by subtracting the additional crossflow diffusion . Some crossflow diffusion has occurred in Huh's scheme. however. Fig. former is conservative while the latter is not. 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.2.11 has a flow given in sections 4.14 show that the analytical solution can be reproduced by Huh's skew differencing scheme for the case.1 and 4. but less than inthe donor cell scheme. It is indicated that Raithby's scheme may give unphysical results in problems with steep gradients of the quantity under consideration.2.138 convection problem. both of them are conservative in a unidirectional flow. 5. 5. The details of those schemes are Figure 5.
13. 5. Comparison c f the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 10x10 meshes along the line CA in Fig.139 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh C A spatial coordinate Fig. 5.3(B) .
5. Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 6x6 meshes along the line CA in Fig.3(B) .14.140 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh 10 5 0 spatial A coordinate Fig. 5.
16 has a flow with 8 = 63.141 constant from the total diffusion constant of the finite difference equation. The corrective scheme is also tested in the problem geometry of Fig.3(B). donor cell solution and the solutions by mesh point aand mesh interface implementations. 5. 5. Figure 5. In Fig. The mesh point implementation gives an unphysical solution . Figure. The full correction of the crossflow diffusion gives the true solution without any numerical diffusion error. 5.19 is a hypothetical 3 x 3 recirculating flow field where the inlet boundary values are specified on the left and bottom surfaces. are introduced in Chapter 4. 5.18 show that the crossflow diffusion can be eliminated by the corrective scheme and that a finer mesh spacing always gives a better solution. although not identical to the true solution. Both of them are tested in a recirculating flow problem with pure convection. 5.17 and Fig.21 gives the true solution. Two implementation strategies of the corrective scheme. The flow is at a 45* angle with respect to the mesh orientation. 5. The corrective scheme gives a physically reasonable solution. Fig. mesh point and mesh interface.20 shows the diffusion corrections of the mesh point and mesh interface implementations at each interface. Figure 5.15 numerical solutions are given for a simple pure convection problem with various diffusion corrections.430. Fig.
3 D=0.99 D=0.0 2.29 5.142 D=0.1 10 8.0 2.25 0 5.88 5 4.0 10 7.41 u=v=l Ax=Ay=l D =D =0.75 6.13 7.0 2.5 x y Fig. Donor cell solutions with various diffusion corrections for a pure convection problem with e = 450 .5 (True Solution) 9.5  3. 5.08 5.71 0.0 1.0 8.88 10 5 0 2.92 5. 0 V ________ D=0.71 5.01 7.5 0 1.29 0.12 5.15.59 7.
73 10 7.07 2.59 1.11 6. True solution and donor cell solutions with and without diffusion correction for a pure convection problem with e = 63.56  2.96 3.41 0.11 0.11 True Solution V 63.93 10 5.16.82 0.10 9.91 0. v=2 x= y=l 2 D =D = x y 3 Fig.430 u=l.07 10 3.07 0.33 1.143 D=0.37 3.430 . 5. 0 D2 3 0.04 4.
5.3(B) .17. Comparison of the analytic solution.144 Analytic solution Numerical solution with donor cell Solution by corrective scheme 10 A spatial coordinate Fig. 5. numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig.
Comparisoin of the analytic solution.145 Analytic solution Numerical solution with donor Solution by corrective scheme 10 5 0 spatial coordinate Fig. 5.3(B) . 5.18. numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig.
146 0 Ofr 3 ~LC~' / 10 1 1 5 0 1 3 3. Simple recirculating flow field for test of the implementation strategies of the corrective scheme . 5. 3 ) 2 3 2 Lx = Ay = 1 Fig.19.
2 0.25 1.75 1.75 1.25 mesh point implementation 0.5 0 0 5 0 0 .147 0.2 0 0 0 mesh interface implementation Fig. 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.20.19 .125 0.75 0.625 0 .5 0. 5.75 0 0 0.625 0.75 1. 5.
21. 5.16 11.83 6. donor cell solution and solutions by the mesh point and mesh interface implementations of the corrective scheme for the recirculating flow field given in Fig. 5.10 8.08 9.148 True solution Donor cell mesh point implementation 5.21 7.94 9.10 9.75 6.16 6.19 .82 8.51 10 5 0 Fig.86 12.21 mesh interface implementation 0.26 8.18 7. Comparison of the true solution.
5. flow field is used here to test the mesh point and mesh interface implementations of the corrective scheme. 0.149 while the mesh interface implementation gives a physically reasonable solution. The true solution cannot be reproduced because some of the information about the flow field has already been lost in Fig. 5. Since the steam concentration distribution and flow field are given at the start of the calculation. 5.2 kg/sec. the steam concentration for the new time step can be calculated in an explicit way.24 in which the source mesh has the highest steam concentration.26.27 is physically reasonable and even better . 5.28 show the steam concentration distributions after one time step. Figure 5. 5. 5. Another test calculation is done for a two dimensional. 5.25 and Fig.25 and Fig. 5.27 and Fig.28 is not a reasonable solution while the mesh interface implementation result in Fig.19. There was initially air in the containment and steam is introduced in the source mesh at the rate of 0. respectively. Figure 5. 5. The steam concentration distribution at the beginning is given in Fig.23.3 sec with the corrections in Fig. After a while a flow field is set up in the This containment by the input mass and momentum of steam.26 show the diffusion constant corrections of the mesh point and mesh interface implementations and Fig. 5.22 shows the flow field in the containment and more detailed information about the flow field is given in Fig. The result in Fig. rectangular containment with an arbitrary recirculating flow field.
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 . 5.150 Ax = 0.0m Fig.22.5m Ay = 1.
0648 1.7126 0.23.2460 0.3074 0.1139 0 .5 m Ay = 1 m Fig.4870 0.3957 0.151 0.3253 0.5174 1.531 0.456 0. 2214 velocity [m/sec] Ax = 0.1276 0 . 045 0.34 0.1422 2148 0.2629 0.2474 04 1379 0.0492 0 . 1317 0.4920 0 . 861 1.0127 0.1385 0. 5955 SOURCE 0.0348 0.3 095 1.2216 0. 3563 0.93 2. 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 .5893 0.207 2. 0735 0.1229 0.0937 0 1.2159 0. 5.2190 0.2435 0.
0136 0.0298 0.0002 0.0358 0. 5.152 0.0676 0.0011 0.1244 0.0007 0.5m by = Im Source: steam concentration in Ikg/m 3 0. Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .1712 0.0095 0.1085 0.0346 0.0179 0.1266 0.24.2 kg/sec of steam Fig.0819 0.0031 Ax = 0.1859 SOURCE 0.0620 0.1592 0.1750 0.0049 0.1178 0.0305 0.1710 0.1420 0.
0590 0.1452 0. 2422 0. 0086 0.0300 0.1 389 0.0311 0.stant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment .25.1694 3106 0.0937 0.1451 0.3262 0..0 183 Ax = 0.C209 0.0253 0. 0. 0227 0.0255 SOURCE 0.0731 116 0.0600 0.1583 0. 0396 0.153 0.1701 0.5m by = Im Diffusion constant in Im2/sec] Fig.1279 0.0 278 0.0 362 0.1178 0.0348 0.0794 0. 0152 0. 0.2740 0.1854 0.0440 0.0224 0.1763 0. 5. 0488 0.0311 0.0193 0.1014 258 0. Diffusion co. 0.
0240 0.4732 0.2134 0.1026 0 SOURCE 0 0 0.0192 0 0.0287 0.0534 0.1149 0 0 0.0 257 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 . 5.0443 0.0632 0.2529 0.0518 0 0 0. 0706 0.9769 0 0 0.0479 0.0 923 0 0.0215 0.3694 0.26.0 111 0 Lx = 0.0860 0.2317 0 0.0 76 0.5m by = lm Diffusion constant in Im2/sec] Fig.154 0 0 0. 0579 0.
5.27.1623 0.1711 0. with donor cell differencing convection term without any diffusion of correction .155 0.0342 0.1363 0.0195 0.0342 0.0465 0.0107 0. 0.0011 0.0003 0.0046 Lx = 0.0947 0.1135 0.1847 SOURCE 0.5m Ly = Im steam concentration in [kg/m 3 ] Source: o.1475 0.2 kg/sec of steam Fig.0735 0.0448 0.0068 0.1321 0.1733 0.0833 0.3 sec.1706 0.1243 0. Steam concentration distribution after one time step.0016 0.0233 0.
1747 0.1797 0.0010 0.0229 0.1646 0.0068 0.1240 0. Steam concentration distribution after one time step.0937 0.0355 0.2086 SOURCE 0.0005 0. 5. 0.0465 0. with mesh point implementation of the corrective scheme .1247 0.0042 Lx = 0.3 sec.0287 0.0250 0.156 0.1911 0.0815 0.1334 0.0092 0.2 kg/sec of steam Source: 3 Fig.5m Ly = Im steam concentration in Ikg/m 0.0433 0.0029 0.0709 0.1729 0.1493 0.28.
0010 0.0039 ax = 0.0424 0.157 0.29.1628 0.0100 0. with mesh interface implementation of the corrective scheme .2 kg/sec of steam Fig.1756 0.1487 0.0815 0.3 sec.0003 0. Steam concercration distribution after one time step.1794 0.0722 0.1738 0.1847 SOURCE 0. 0.0481 0.0228 0. 5.0295 0.0947 0.5m by = Im steam concentration in [kg/m 3 Source: 0.0192 0.1112 0.1243 0.1310 0.0007 0.0058 0.0360 0.1342 0.
In Fig. Figure 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.32 shows the profiles of 4 at the same loca tion for the time step size of 1 sec. it has the maximum time step size that can give a physically reasonable solution. 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.3 ADI Solution The ADI scheme is tested to use a time step size larger than the Courant limit.31 shows the profiles of 4 along the dotted line for the time step size of 0. =10.30. 5. The ADI and explicit schemes give almost identical results during all the transient.5 sec. In the meshes at the top of the source mesh.30 there is an inflow from the bottom surface at the velocity of v=1. Figure 5.0. which is exactly the .158 than the donor cell solution. 5. The results of the ADI and explicit schemes are compared in the problem geometry of Fig. 5. 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. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable.0 and the inlet boundary value of 4 is. when the Courant limit is 1 sec.
0 Ax = Ly = 1. v = 1.0 on this line are compared.YXI~^.~ ~ILI_~_~LI*L~l*IIIU*XI 1 9 IB Source of ' 'A Profiles of 4 = 10.1I I _~ 1_. Problem geometry for comparison of the explicit and ADI solution schemes .30.0 Fig. 5.0 c= 1.
01 12.0 10.0 12. Comparison of the explicit and ADI solutions for along the line AB in Fig.0 11.0 12.5 sec 12.30 the profile of sec and the Courant limit of 1.5 . 0A Explicit ADI 10.0 11.0A 12.0 11.0 11. 5.31.0 A iatlai B 10.0 10.0 10.0 11.0 Fig.0 sec for Lt=0.0 coordinate 11.160 12.0 0. 5.0 10.
0 12.0 11.32.0 ~ spatial coordinate 12.0 11.0 10.0 4 sec 11.0 11. 12. 5.161 1 sec 12.0 10.0 sec .0 12.0 Explicit ADI 10. 5.0 16 sec 20 sec 12.30 for At=1. 11. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig.0 .0 10.0 Fig.0 sec and the Courant limit of 1.0 10.
5. but it damps out rapidly and the ADI scheme gives almost identical results with the explicit scheme. 5.32 and Fig. Fig. . there is some deviation between the two results.31. 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. Fig.33. Initially.33 shows that the ADI scheme can use a time step size up to five times the Courant limit. 5.162 the Courant limit. No crossflow diffusion corrections are involved in the results of Fig. 5.
0 At =0.0 t spatial coordinate B Fig. Comparison of the ADI solutions for the profile of at t=10.0 sec along the line AB in Fig.33.163 12. 5.5 sec ec 11.30 with different time step sizes . 5.
or superheated steam as may be appropriate. in the finite difference donor cell treatment of convection. The total diffusion con stant is the sum of the laminar and turbulent diffusion constants.2 Numerical Schemes (1) There are two numerical diffusion sources. and turbulence equations are solved separately using that flow field. truncation error and crossflow diffusion. Crossflow diffusion occurs due to the donor cell treatment of the convection term in a multidimensional problem. energy.164 CHAPTER 6 CONCLUSION 6. The thermal equilibrium is assumed and the phase change occurs to maintain 100% relative humidity. The mass diffusion.1 Physical Models (1) The solution scheme presented in Chapter 2 has been adequate for modelling the slow mixing stage in the containment after a lossofcoolant accident. (2) The models of laminar and turbulent diffusions in Chapter 3 may be adequate for predicting the hydrogen transport in the containment. 6. The . The con tinuity/momentum equations are decoupled from the scalar transport equations and solved by the SMAC scheme to obtain the flow field.
most of the numerical diffusion error in a multidimensional. have been tried in order to eliminate the numerical diffusion. convection dominant. The skew differencing scheme gives good results for some problems. skew differencing and corrective schemes. the corrective scheme is generally preferred to the skew differencing scheme. It is conservative and an explicit scheme can be used without affecting the simple solution structure. It gives unphysical results for most recirculating flow and coarse mesh problems. ~unx. The inclusion of corner points complicates the matrix structure and a fully implicit scheme should be used for maintenance of stability. The corrective scheme is based on the fact that the additional crossflow diffusion can be predicted theoretically for every mesh at every time step. (2) Two types of approaches. but not always. The truncation error diffusion occurs in the flow direction while the crossflow diffusion occurs in the diagonal direction of neighboring mesh points. 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. Therefore. recirculating flow problem is due to the crossflow diffusion. . Therefore.165 effective diffusion constants of the truncation error and crossflow diffusion are of the same order of magnitude.
First. using the characteristics of a quadratic equation. Implicit (ADI) scheme can be used to increase the time step size and decrease the overall computational effort.4. The graphical method is used to obtain the stability condition in terms of the Courant. mesh point and They mesh interface. i. are identical in a unidirectional flow. dx and Rx . Cx. 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.e. (4) The Von Neumann stability analysis is applied to various finite difference forms of a general conservation equation.166 (3) Two implementation strategies. the physical constraint in section 4. diffusion and cell Reynolds numbers. are tried for the corrective scheme..2 imposes a maximum time step size that can give a physically . (5) The maximum time step size is limited by the Courant The Alternate Direction condition in an explicit scheme. The results of various sample problems show that the mesh interface implementation of the corrective scheme always gives a physically reasonable solution with negligible numerical diffusion error. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable. it has its own limitations due to the following factors.
Second. For example.167 reasonable solution. . the ADI scheme is a fractional step method. asymmetry of the given problem may give different results according to the sweeping sequence. a steadyThe state solution may never be reached due to the fact that one time step is composed of a few fractional steps.
The model should show proper limiting behaviors as the Reynolds number goes to infinity (turbulent) and zero (laminar). .g. A criterion for the flow regime should be developed in such a form that it may be implemented into a computer code. etc.168 CHAPTER 7 RECOMMENDATIONS FOR FUTURE WORK 7. The flow regime criterion may be replaced with a model that covers all the turbulent.2 Numerical Schemes There have been two approaches. They may still be improved further to get an accurate solution in a transient. it is necessary to predict the flow regime in order to use appropriate physical models.1 Physical Models Since both the laminar and turbulent flow regimes are possible in a complicated geometry and flow field. 7. The modified kE model in a low Reynolds number flow may be a guide in this approach. previous time step information. The important parameters may be the velocity. geometry (e. velocity gradient. multidimensional recirculating flow problem. time step size. skewdifferencing and corrective schemes. distance from the wall).. transitional and laminar flow regimes. to eliminate the crossflow diffusion.
ADI and implicit schemes although only the implicit scheme has been used up to now. 2. Skew differencing scheme 1. It may be possible to develop a reasonably simple form that may or may not include the corner points. . The stability of the skewdifferencing scheme is open to question in the explicit.169 Corrective scheme The corrective scheme has been successful by the mesh interface implementation strategy in the scope of this work. More tests and validation calculations are required in recirculating flow problems. The solution in Appendix A may be a useful guide in this work. The conservative form of the skew differencing scheme is complicated and time consuming in comparison with the corrective scheme.
. and Harlow. 5473 (1980).H. 24. and McDonald. Phys. 6. "Hydrogen Distribution in a Containment With a HighVelocity HydrogenSteam Source. Bradshaw.User's Manual.. A." Journal of Comp." Journal of Comp. E.. and Harlow.R. [2] [3) [4] [5) [6] [7] [8) [9) [10) [11) [12] . Briggs.A. B.R. D. "Solution of the Multidimensional Compressible NavierStokes Equations by a Generalized Implicit Method. A... 94110 Amsden.H. thesis (1979). H. Stewart. et al. Bloom.. "On the Structure and Use of Linearized Block Implicit Schemes.A. Phys. 3." MIT Ph. W. Methods in Applied Mech. Briley.V." Numerical Heat Transfer 3.A..R. "A New Finite Element Formulation for ConvectionDiffusion Problems. "A Finite Difference Scheme for the Incompressible AdvectionDiffusion Equation. and Eng. Engineering Calculation Methods for Turbulent Flow." HEDLSA2682 (October 1982). Phys. T. and Whitelaw. 34. Comp.." NUREG/CR1148 (EGG208).R.R. Briley. J. W.R. Broadus. Transport Phenomena. and McDonald. "BEACON/MOD 3: A Computer Program for Thermalhydraulic Analysis of Nuclear Reactor Containments . 372397 (1977)." Journal of (1968). W. and Lightfoot.. "The SMAC method: A numerical technique for calculating incompressible flows. (April 1980).. Boyle. John Wiley & Sons (1960). D." LA4370 (May 1970). "Transport of Turbulence in Numerical Fluid Dynamics. S. Cebeci. "Transient Effects in Turbulence Modelling. 393409 (1980).B.N. "A Simplified MAC Technique for Incompressible Fluid Flow Calculations. Amsden. et al. H.D... Phys. A.170 REFERENCES [1] Amsden.E.. Bird. Baliga.H. 6. and Harlow. 233241 (1975). and Patankar." Comp. 332325 (1970). R.H. G. C. F.G." Journal of Comp. F. F.. Academic Press (1981). P.
F. J. 1344 (1981)." The Phys. 471484 (1967). Y.H. T. "An Evaluation of Upwind and Central Difference Approximations by a Study of Recircul ing Flow.S. Hirt." Computers and Fluids. Chang. Williams. M.D. thesis (1977). P. and Hansen. 95107 (1982). Keller. Cloutman. P. 127137 (1982).. N. and Ramshaw. 1. "Tensor Viscosity Dukowicz. Cook. and Nakayama. De Vahl Davis. Phys. 7179 (1979).171 [13) Carbajo. G.D. K." MIT M. Method for Convection in Numerical Fluid Dynamics. 32. G.. B. Chang. Vol. Crowley.. 27. Methods in Applied Mech." MIT Ph." Nuclear Eng.H.I.A Computer Program for Calculating Timedependent Turbulent Fluid Flows with Slight Density Variation." MIT Ph. S.. Covelli. Phys. "SOLAICE: A Numerical Solution Algorithm for Transient Compressible Fluid Flows. "Accuracy of Difference Formulation of NavierStokes Equations. Varadi." Comp. 2943 (1976). and Lewis.. "SecondOrder Numerical Advection. G. Chen. W.D." Journal of Comp. "Study of Nonhomogeneous Conditions in Reactor Containments Due to Separation Effects." Nuclear Eng...G. S. and Mallison.K. 4." LA6236 (July 1976). [14] [15) [16] [17] 118] [19] [20) [21] [22] 123] [24] [25) .D.I.P. Cebeci. "Comparative Analysis of Numerical Methods for the Solution of NavierSTokes Equations....B.. thesis (1980). 1 (Nov.. L..J. R. of Fluids Supplement 2 (1969). Hirsh.D. "Numerical Methods of Reactor Analysis. "Studies of Numerical Methods for the Plane NavierStokes Equations. "VARR2 . and Design 69.L. Nielsen. M. Cheng. Clark.B. Jr. "Buoyancy Effects in Mathematical and Numerical Turbulence Modelling. thesis (1981)... Vol. S.W. Academic Press (1964).. "Simulation of Containment Cooling with Outside Spray After a Core Meltdown. 1976).S..." CRBRPARD0106. and Romero.C. J. and Design 69. C. J. L. "Coolant Mixing in the LMFBR Outlet Plenum. and Eng.. J." Journal of Comp. H..
Curtiss.E. Turbulence. and Marohar.H. C.. Wiley.The 1978 Freeman Scholar Lecture.." Journal of Comp. 528 (March 1979)." Int. F. New York (1954).E. Harlow. C. T. R. J.. J.F. 18." The Phys. B..B. 8093 (1968).. "A Comparison of Hybrid and Quadratic Upstream Differencing in High Reynolds Number Elliptic Flows. Phys.O. 1952 (1975).. "Behavior of Buoyant Moist Plumes in Turbulent Atmospheres. Humphrey. G. Hamza. and Launder. "A Numerical Fluid Dynamics Calculation Method for All Flow Speeds. Phys. McGrawHill Book Company. 2.... B. 8195 (1981).A. Harlow. "Improved Form of the Low Reynolds Number kc Turbulence Model. F. [38) . Han. 339355 (1968). 197213 (1971).." Journal of Comp." MITEL 81020 (June 1981).A." Journal of Comp. Gupta. of Fluids Supplement 2 (1969).W.O. Phys.. of Fluids. 3. and Golay. Hirt. "Heuristic Stability Theory for Finite Difference Equations. M." Journal of Fluids Eng. (1975). "Computer Studies of Timedependent Turbulent Flows. and Amsden.W. J. M.. Molecular Theory of Gases and Liquids... 2nd ed. "On the Use of Central Difference Scheme for NavierStokes Equations.172 [26] Fromm.H.C. R. and Amsden.. 309312 (March 1975). for Num. Vol. [27] [28] [29) [30) [31] [32) [33) [34] [35] [36) [37) Hirt. "Buoyancy Induced Fluid Motions Characteristic of Applications in Technology . and Eng. A. J. 29. C. A.. 15.H." The Phys.M. Harlow." The Phys. 101. Vol." Comp.. and Bird. Phys.H.A. 17.W.. 557573 (1980).. Hinze. Hoffman. "Numerical Calculation of Multiphase Fluid Flow.P. "Numerical Calculation of Almost Incompressible Flow. F. 8. A. J.. Methods in Eng. "Practical Investigation of Convective Difference Approximations of Reduced Dispersion. Hirschfelder. of Fluids Supplement 2 (1969)... Gebhart. Vol. and Amsden. Methods in Applied Mech." Journal of Comp. R.A.
part 3. "Analytical Modelling of Hydrogen Transport in Reactor Containments. and Golay..P." J. "The Prediction of Laminarization with a Twoequation Model of Turbulence. Heat Mass Transfer. V.E. and Rodi. "The Numerical Comparison of Turbulent Flows..173 [39) Houston. and Bremhorst. D. F. Keenan.A.J. G.H. 221239 (1974). 103. V. "Experimental Investigation of Hydrogen Distribution in a Containment of a Light Water Reactor Following a Coolant Loss Accident. "A Modified Form of the kc Model for Predicting Wall Turbulence. "Progress in the Development of a Reynolds Stress Turbulence Closure. and Launder. 3. and Spalding. Launder. Cl." Comp. "ADI Solution of Free Convection in a Variable Viscosity Fluid.P." BF63. Steam Tables. Langer. [40) [41) [42) [43] [44) [45] [46] [47] Leschziner. 1981). Methods in Applied Mech. Launder.... 16. Phys." Trans. 301314 (1972)." Int.3633... of the ASME. B.W.P. B.E. Vol. 68.. K.G.. M.B. 23.. C. 16.G. Jr. 11191130 (1973). John Wiley & Sons (1978)..E. 293312 (1980)." Report submitted to 2nd International Topical Meeting on Nuclear Reactor Thermalhydraulics (March 1982). Battelle Institute e. W. "Practical Evaluation of Three Finite Difference Schemes for the Computation of SteadyState Recirculating Flows. et al. Reece. Vol." Comp. "Analytical Modelling of Hydrogen Transport In Reactor Containments." MIT Sc. thesis (1983).. G. "The Calculation of Low Reynolds Number Phenomena with a Twoequation Model of Turbulence. 148) [49) . 269289 (1974)..D. Jones. B.. W. B. and Keyes. 537566 (1975). and De Braemaecker. J. and Eng. Frankfurt. and Launder. and Eng." Int. J.. Jones.V. M. W. Manno.E. translated (October 1980). Heat Mass Transfer. Fluid Mech.. J." Journal of Comp.. Vol. Vol.H.K. Lam. J. 15. Methods in Applied Mech. M.P. 456460 (Sept. Manno.
. and Sherwood. 9.. G. of Fluids Supplement 2 (1969). McGrawHill. Hermosa Publishers.D. "A Note on the Empirical Constant in the KolmogorovPrandtl EddyViscosity Expression. Methods in Applied Mech. P. for Num. 153164 (1976).B. John Wiley' & Sons." Int. The Properties of Gases and Liquids. and'Eng.. "Upstreamweighted Differencing Schemes and Their Application to Elliptic Problems Involving Fluid Flow. Patankar.B. [52) [53) [54) [55) [56) [57) [58) [59] [60) [61) [62] [63) .. Albuquerque.C. 169184 (1972). and Weiss. K. Roberts." Comp. Roache. 2.J.. 15. and Spalding. 20 (1966). W. "A Critical Evaluation of Upstream Differencing Applied to Problems Involving Fluid Flow. G. R. 9. Vol. S. N. G.O. Mass and Momentum Transfer in Three Dimensional Parabolic Flows. S. 191206 (1974). 10. and Eng. Computational Methods in Engineering and Science. "Convergence and Accuracy of Three Finite Difference Schemes for a Two Dimensional Conduction and Convection Problem. S. Rodi.174 [50] [51) Nakamura." The Phys.. P.. Raithby.. (1977). Methods in Applied Mech." Int.V. New Mexico (1976).... D. Patankar.V. "SkewUpstream Differencing Schemes for Problems Involving Fluid Flow. 386389 (Sept.J.D. Heat Mass Transfer. Raithby. 1975). 75103 (1976). Rodi." Journal of Comp. "On Artificial Viscosity.K. Schonauer. J." Trans. Runchal.K." Comp.. "Numerical Experiments with a Difference Model for the NavierStokes Equations (Turbulence Model). K." Computers and Fluids.. A. of the ASME. 541550 (1972).. 17871806 (1972).. Hemisphere Publishing Corporation (1980).. Computational Fluid Dynamics. Inc.E. Pergamon Press (1982). Roache.D. W. Raithby.. J.." Mathematics of Computation. T. Methods in Eng. "Convective Difference Schemes. "A Calculational Procedure for Heat. Phys. 4. New York (1958). Turbulent Buoyant Jets and Plumes. and Torrance. Reid. W. Numerical Heat Transfer and Fluid Flow.
E.. G. Strong.S. Phys. Stuhmiller. "A Novel Finite Difference Formulation for Differential Expressions Involving Both First and Second Derivatives. "Proposal for a New Discrete Method Based on an Assessment of Discretization Errors." Int. A First Course in Tur [65] [66) [67] [68) [69) [70] [71] Torrance. A. Trent." Comp. G. Y..A. Methods in Eng. Budden.. 551559 (1972). and Lumley. Methods in Applied Mech. D.B. Spalding. L. (Jan..W. 1968). Raithby. Wilke.. K. Vol. "Development and Validation of a TwoVariable Turbulence Model." SAI74509LJ. 1981). "TEMPEST: A Threedimensional Timedependent Computer Program [72] for Hydrothermal Analysis. and Valentin... R." FATE80114 [73] (Jan. and Woolner.. 411428 (1980). No. "Simulation of Convection and Diffusion Processes by Standard Finite Difference Schemes and by Influence Schemes. M. 4. 35." Journal of Comp. A." Numerical Heat Transfer. and Strong. 411428 (1980). G. 3.Dec. "The Numerical Treatment of Advection: A Performance Comparison of Current Methods. 20. Tennekes. A.. Stubley.D." Chemical Engineering Progress. J. 46. C. bulence.B. 4 (Oct. Raithby. . MatheResearch matical Sciences.. 72B." Numerical Heat Transfer.M.175 [64] Shin.R. Stubley. Smith. R.D.B.G.. J. 220237 (1976). For Num. MIT Press (1970). No.. H.L. K.B. D. Vol.D.J.." of the National Bureau of Science .. 95102 (1950). "Comparison of Finite Difference Journal of Computations of Natural Convection. 153168 (1982). and Hutton.L.D. "Diffusional Properties of Multicomponent Gases.H. and Eng..A. G.. 2. "Numerical Analysis of Fluid Hammer Waves by the Method of Characteristics. 3. J.. 1974). and Eyler.
2 is simplified further by substitution of 4 for T as follows.1. a= k PCp C =2 u 2A.176 APPENDIX A ANALYTICAL SOLUTION FOR THE PROBLEM IN FIGURE 5.1) In Eq.3) ax ..3(B) and the constants can be grouped into one constant. A.a2T aT 2 + a2 2 2c ax ax 2 y ay (A.1 convection occurs in the x direction and diffusion occurs in both x and y directions. Eq.3(B) A steady state two dimensional energy conservation equation without any source is given in the follbwing Eq.1 is2 reduced to the following form.2) Equation A. Therefore. 2 x 2 a2 ay = c 2 (A. c. The boundary conditions are specified in Fig. pc u aT = k(a2T + 2aT 2 x2 p x ax ay (A. 5. as follows. A.
Equation A.a =  (A. ables.8) dy where a > c > 0 . c 2 X cx 2 = (A. The variable 1 is separated into two vari X and Y. tively. A.3 is reduced to the following form.3 can be solved by the method of separation of variables. which are functions of x and y only respec (x. A.6 should be equal to some constants because they are functions of x and y only respectively and the sum of them is equal to a constant.7) d Y c .4) The boundary conditions should also be expressed in terms of 4 as given in Fig.177 where 4 = eCXT (A. I d2X + 1 d 2Y 1 = c2 C (A.y) = X(x)Y(y) (A.6) dx dy The two terms on the left hand side of Eq.1. A.5) Then Eq.
4 = ec T . Boundary conditions in terms of 4 for the problem in Fig. A.1. 5.3(B) where 0 is given by.= YO T 0 ii ii i h YO + co = 0 4 = 0.178 .0 0 = 0 x Fig.
7 and Eq.10 should satisfy the following relations. A.8 can be easily obtained as follows. C3 = 0 (A. The boundary con dition at x=0 will be considered later.1 may be given in terms of the variables X and Y as follows. dY dy y=O Y=O 0 (A. A. the constants in Eq.11. 13. A. A.12 and Eq.12) (d dx _cX) = 0 x=xO (A.16) . Eq. A.9 and Eq.11) dY dy=Y 0 = 0 (A.9) Y = C3sin8y + C4cosSy (A. A.179 The solutions to Eq. X = Cleax + C2 eax (A.14) (A.10) The boundary conditions in Fig. A. 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.
the constants A.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.2.19) .y) = e cx (x.y) where nr (A. Th 0 2. .0 2 A n Th h nr 1 n 1+( a c )e n 1 a +c 2a x . 17) T(x.) (A. Therefore.3 . . the final solution can be expressed as follows. nw sin2 nO0 (n=1.1 are determined as follows.180 where 2 =c 2 + n (O) n 2 yO =c 2 + 8 2 .
When the tilde phase of the SMAC scheme and other scalar transport equations are solved by the ADI scheme. Equations B. there are six sweeps in a three dimensional problem and four sweeps in a two dimensional problem. The most important sweep should be given the first priority and most updated information. Since there are two steps involved in solving the momentum equation in each direction. They are formulated for a natural convection problem where the gravity acts in the zdirection. The sequence of the sweeps is arbitrary and depends on the problem under consideration. the computational effort for one time step is greater than that of the explicit scheme. 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.1Eq.1 and B. The time step size limitation in the ADI scheme comes from the semiimplicit treatment of the nonlinear convection term. The ADI scheme is applied to a two dimeraional momentum equation in the following Eq.i. 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.3 are the z and x direction sweeps of the x direction momentum .4.iillaY~i~~l_ ^YII^^~II ( . Therefore. B.
ar .2 and B.2) n+l i+j * .1) * * n * *  ij+ At Wi+1 j+ ij+ + i+Ij+wij+ uij+Wilj+ 1 P n~ .182 equation.ui+ j At  n+l * u i+lj i+ .2ui+.u n+l ui+2 2 n+l Ar2 +2u j n+l ui j .Rxi+ jui+ jui+ j RX * n+l (B. Equations B.Wi+ Az j Ui+ j 1 ui+ j+l * . 2 Az + Ui+ 1 P  n  i+j i+ jui+4j (B.4 are the x and z direction The z direction sweeps of the zdirection momentum equation.2ii+ Ar 2 + Wilj+ . u n * u j i+ Ui+D Dt j + * n i+ j+ Ui+ j .u Lr * n+1 .P z  n * RZij+ ij+ ij+ (B. sweep of the z direction momentum equation is treated as the last one in order to be given the most updated information.3) .
4) Equations B. B.183 n+l i+ * wij+ * i n+l +lWij+ * n+l zW.1.3. .4 have given almost identical results with the explicit scheme for a time step size below the Courant limit.+ n+1 n+l Az 2wi + w _P z P0 zP g z RZ n+l ij + wij+ 3wij+ (B. w At n+l Wi. B.2. and B.
4.8. y and z directions. a flow split occurs as shown in Fig. is the program for testing Fig. 5.3(A). 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. 'prog2 fortran' and 'prog3 fortran'. The file. y and z direction mesh interfaces.2. are the computer programs for calculating the numerical values of the analytical solution in the problem geometries. the calculational logic of the mesh interface implementation of the corrective scheme in section 4.3. 'progl fortran'.8. 'difprg fortran'. 5.184 APPENDIX C COMPUTER PROGRAMS The files. When there are two outflows in any of the x. There are When there are inflow and outflow.3(B) and Fig. the crossflow diffusion constants can be calculated in a simple way as in the first part of the program. are assumed to be zero. Then four crossflow diffusion constants at every interface are summed up to give a total crossflow diffusion constant . two different cases treated separately. 5. Fig.. 4. or all inflows in x. All the inflow velocities The component velocities are used to calculate the crossflow diffusion constants for eight subspaces in a three dimensional coordinate space.
185 because one axis direction. A portion of the VARR program is introduced to show how the ADI scheme is implemented in the energy and momentum equations. 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. The subroutine 'tdm' solves tridiagonal matrix problems by forward and backward sweeps. The ADI scheme for the momentum equation is given in Appendix B. has four neighboring subspaces. . the positive x direction.
OALNOEXP(ALN.O) ALNX .14 55 ATOT * THIODEXP(ALN*XO)/2.0 44 YY I IIPI*(UJ0.NY. AAN*2.YO) CCOrrF 2.2) PI a 3.OZ) T(tO) DIMENSION REAO(10.NY  0.Tl.EO.N FILE: PROG FORrRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT REAL*0 (AH.0 00 20 IPsi.215.U.O*ALNXO)/(ALNC) COF * CCOFF/lI AAN .ALP II FOIRMAT(F7.5F11.26 II IF  IP  I GO TO 55 ALN = DSORT(CSO*II*II*YSO) (II.1 ) DX.COF*DSIN(II*HPI)/(1.NX.100) (T(I).4ALNX) ATOT .5)/NY ADD * ATOTTDCOS(YY) SUM T(d) v SUM + * ADO * DEXP(C*XO) 20 CONTINUE SUM 10 CONTINUE WRITE(G.2.3) SlOP END PRO00210 PRO00220 PROO0230 P1ZOOO240 P1I000250 PROOO260 PROOO270 PROOO200 PROOO290 P11000300 PROO0310 PRO00320 PRO00330 PROOO00340 PRO00350 PROOO360 PROOO370 PROOO300 PROOO390 PROOO400 .IALNIC)*DEXP(2.000190 PI00O200 NY*OX YSO v PI*PI/(YO.I=l.0 IHPI DO SUM 10 JJI.1415926536 C = U/(2.0*TII/PI PI/2.04ALP) CSO .3F7.NY) 100 FOnMAT(IOX. C*C XO " NXOX YO * PROOOO0 PROOOO20 PROOO30 PROO0040 PRO000OOOO P14000070 PRO000OGO P11000090 POOOOT100 PROOOO10 PIIOO0120 PRO00130 PROOO140 PROOO150 PRO00160 PROO000170 PRO00100 P1.XO)/(ALNC) GO TO .
U.5)*DX = YC = (JJ.+ALNX) BEN 11*PI/YO YOO .0 SUM3X 0.0 .p A FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 MPI.5+YO 00 20 IPs1. 1415926536 C .OCOS(OEN*Y4) Y4S DOSIN(3EN'Y4) ADO * AAN*(EXALO + ALNX*XALO)*YOO PROOO520 PR000530 P1OOO00540 PR000550 .DSIN(BEN+Y3) Y4C .0 = 0. NX.3F7 .DCOS(BEN*YI) YI Y22 • DCOS(BENY2) OCOS(BEN*Y3) Y3C Y3S .NY2 0.XN = XN . GL(10). GH(10).2) PI .5*YO .11) DX. 11 FORMAT(F7.215.(ALN#C)*DEXP(2.0  RTO DO =  OSORT(XXX) 10 JJ=1.X = 1. ALP READ(10.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN .OALP) CSO = C'C XO * PROOOO O PRO00OO20 P R000030 PROOOO40 PROO00SO POOG0050 PROOOO000 PROOOOOO PROO0090 PIR000 100 P11000110 PRO00 120 PROOO 1.1.FO.0 SUMO SUMI r 0.2.0 SUM2 = 0.O) GO TO 55 ALNX .0 . TL(10).5) EXAL = OEXP(ALNOX) XAL = DEXP(ALN'X) EXALN D EXP(ALN*XN) XALN = DEXP(ALNXN) IF(II.COF*DSIN(II*HIPI)/(I. NY.0 SUMaX SUM4 XN vI Y2 Y3 Y4 • 0.21 II = IP  I ALN = DSORT(CSO + IIII*YSQ) EXALO = DEXP(ALN*XOO.OTHIH/PI 11PI = PI/2.30 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. IC1l REALi0 (AII.U/(2.5) XALO = DEXP(ALN*XO*O.0 X = (JJO. TC(IO) TH14.DCOS(BEN*YC) .0 NY2 XXX NY/2.0Z) DIMENSION TII(10).O.3.0 = YO/2.21)*DX = X 4 YO/2.0 X + XO/2.0 SjU3 " = 0.
SFII. 1OO) (TH(I).0 ADOvY * 0.I WRITE(6.EXALN + ALNX*(CALN)*XALN)*Y3C EN*AAN#(EXALN .DEXP(COXN)/RTO PROO000560 PROO0570 PROOOS80 PROOO590 PROOO600 PROO0610 PROOOG0620 PR000G30 PROOO640 PRO0650 PRIOOOG60 PROOG0670 PROOO680 PROOOG90 PROOO700 PROOO7 O PROO0720 PROOO730 PROO00740 PROOO750 PROOO7GO 1PI000770 PROOO700 P000790 PROO000000 PROOOIO1 PIOO000120 PROOO030 PROO0040 PROO000050 PROOO060 PROOO870 PROO000 PROOO090 PRO00900 PROO0910 PROOO920 PRO00930 PR000940 I0760 00 oo I .tOO) (TC(I).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.(" % FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 ADOI w AAN*(EXAL+ALNX*XAL)*YII AD02 .100) (GH(I).AAN*(EXALALNX*XAL)*Y22 ADO3X v AAN*((C+ALN).I 1OO FORMAT(IOX.5 A)DO * TIHl'XALO*O.5 A002 * TMIIIXAL0.NY2) WRITE(6. ALNX*XALN)*Y3S ADO3Y * ADD4X * AAN*((C+ALN)*EXALN + ALNX*(CALN)*XALN)*Y4C + ALNX*XALN)#Y4S AIDD4Y * EN*AAN*(EXALN GO TO 77 55 AOOI = TIIH*XAL*O.II.0 AOOX a 0.NY2) = t.00) (GL(I).0 ADO3Y * 0.NY2) WRITE(G6.II.NY2) WRITE(.5) SUMI*DEXP(C*X) SUM2*DEXP(C'X) (SUM3X+SUM3Y)*DEXP(C*XN)/RTO (SUM4X+SUM4Y).3) STOP ENO SUMO*DEXP(C*XO0O.tOO) (TL(l).II.5 ADO3X 0.
NX.S)#(XOYO/(2.C P11000220 PROO0230 PROO0240 P1000250 PROOO2GO PROO0270 P11000200 PROO290 PR000300 PROO03 10 P 11000310 PROOO320 00 PRO00330 PR00OO340 PROO0350 00 20 IP=1.02) 111(10).0 SUM3X = 0.DCOS(iEN*tY2) YV3C = DCOS(BEN'Y3) Y3S = DSIN(BENtY3) YAC = OCOS(3EN*Y.0)/NY2 X4XOYO/(2.5)*YO/(2.0 NY2 nRo P000O070 13100000 PROOOOOO PROOO0000 PRO00110 = NY/2.0 .0 SUM3Y = 0. ALP PROOOO10 GL( O).5)*YO/(2.RT3)+(dJO.OtTIIIM/PI HIPI = PI/2.0)/NY2 X2=(JJ0.5)*(XOYO/(2.0(JJ0.5) * (YO/3.O*RT3)+(JJO.5)*(XOYO/1r3)/NY2 YC=(JO.SO5T(2.0+(dJO.0*RT3)+(J0.S)*(XO/RT3YO/6. TC(1tO) PROO0020 PROO0030 PROO0040 PnOOO050 PROOG000 READ( 10.O*NY2) .5)YO/(2 .22 II IP I PR00360 PRO00370 PROOO3UO P(1000390 P000400 ALN = DSORT(CSO + II*laIYSO) GO 10 55 Ir(II.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN = COF*OSIN(II'HPI)/(I.(dJ0.0 XC=Y0/(2. TIIl.0 SUM4X = 0.o) ALNX = (ALNwC)tDEXP(2.215.0*ALP) CSOQ C*C XO = DX'NX YO = DX*NY YSO = PI*PI/(YO*YO) CCOFF = 2.0) DO 10 00=1.5)YO/(2. G1(10).2. NY.1415926536 C = U/(2.0 SUMI = 0.NY2 SUMO = 0.0*RT3))/NY2 Y3=(JJO.0.+ALNX) OEN = II'PI/YO YOO .0) PROO000120 PROO000130 PROO0140 PROOO150 PR000O160 RT3 = C SORT(3. ( 10).0*RT3#NY2) Y2=YO/2.0.EO.*0RT3#NY2) Y4=YO/3. I 1) DX.NX*DEXP(ALNXI))*YI I ADD2AAN*(DEXP(AI.0 PR00O170 PR000100 PROO0190 PR000200 PlO00210 SUM2 = 0. FILE: PROG3 FORIRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT DIMENSION C 11 REALtO (AH.5)*(XO/RT3YO/6.0*NY2) X3YO/(2.0*RTJ))/NY2 YIYO/2.2) PI = 3.DCOS(UENOYC) YIl DCOS(OEN'YI) Y22 . U.N*X2)+ALNX')OEXP(ALN*X2))*Y22 AD003XwAAN((C+ALN)*DEXP(ALN9X3)4ALNX*(CALN)*DEXP(ALNtX3))Y3C PRO000410 PROO420 PR000430 1R'000440 P1100050 PR00O4GO PROO0470 PlOOO400 POO11000490 PROOOSOO 1100005 10 PROOO520 P0000520 P1OO0530 PROOO040 P1000550 . TL(IO).0 SUM4Y = 0. rORMAT(F7.O' XO/RT3)/NY2 XI=(dJ0.XO/RT3.C.4) Y4S = DSIN(6ENlY4) ADDO=AAN*(DEXP(ALN*XC)4ALNX*DEXP(ALN*XC))*YOO ADDIAAN9(DEXP(ALN*XI)+AI.3F7.
5*RT3)*DEXP(COX3) GHi(JJ)(SUM4X#O.100) (TC(I).100) (TH(I).5+SUM3Y*0. = 0.NY2) WRITE(G.100) (GL(I).I1.OEN*AAN(DEXP(ALN*X4)4ALNX*)DEXP(ALN*X4))*Y4S GO TO 77 55 ADDOTIIII*DEXP( ALN*XC)*O.1OO) (TL(I).3) STOP ENO .0 ADDOX a 0..5 ADDH11H*DEXP(ALN*XI)*0.5 ADO3X PROOOSGO PROO00570 PR1000500 PROO000590 PROOOGOO P1100000.0 77 SUM1 .0 ADO3Y = 0.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.I= .SF11.5 ADD2TtII#DEXP(ALN*X2)*O.5*RT3+SUM4YtO. PROOO6tIO .NY2) tOo rORMAT(IOX. 100) (GI(I).NY2) WRITE(6.I=.I=I.NY2) WRITE(6.5)*DEXP(C*X4) 10 CONTINUE WRITE(6.11.X4) ALNX*(CALN)*OEXP(ALN*X4))*Y4C ADO4Y.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.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.0 ADO4Y = 0.
O. bt ts .DIrn*ISI .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 .WW4DELTII#( PPX + PPZ )/PSUM C RiFGII .0) IF(VM.0.0 ) C DIFR .DIFII'IS2 rGR RfGTI .0 IF(UM.VELT(O).O 15S3 .0 RFGCR 0.DIFRA(0)1.5) DELR.IFZA(0).GT.LT.O 151 4 0 ISS1 O 156 O DIFOOOIO DIFOOO20 DIFOOO30 OIF0040 011:O0050 DIFOOOGO DIF00070 DIFOOOTO DIF00110 RFGRI = 0.UM.GT.5.VELZ(0).2) UU * UI + U2 VV .O.O.0) IF(VP.0 OMT .WW/DELTII PSUM .4 ~ FILE: O FI'IlG FOITIAN A VM/SP COINVERSATIONAL MONITOR SYSTEM PACE 001 DIMENSION VELX(0).2.0) IF(UP.ISC.I 01o00200 DIF00210 DIF00220 DIFOO230 DIFOO240 DIF00250 DIF00260 DIFO0270 DIF00200 OlFOO290 DIFOO300 DI F003 10 I'I .O.0.) ISA .WP 1SI * 0 152 .t = 1 = i v 1 * .VP.UU*DELR*( PPZ + PPT )/PSUM DIFZ * VV*DELZ*( PPX + PPT )/PSUM DIFT .WM.EO.0 n I FOO 130 OMTI m 0.EO.IS .0.5*( VP WI * 0.DIFZ'IS3 * DIFZ*IS4 RFGr GO TO 00 AOS(UM) + ABS(UP) DI F00170 DIFOO IGO DIFOO00190 DIFOOlOO ISI IS2 IS3 IS4 155 ISG .O.UU/DFLR PPZ * VV/OELZ PPT .DELZ.5*( VM V2 0.OR.DEL'TII.VM.5+( UM U2 0.0 RFOT 2 0.DIFTA(0) REA( 10O.O.OR.LT.EO.EQ.ISt + IS2 ISO IS3 + IS4 ISC ISS5 + ISG C UI = 0.5*( WP C IF( ISA.2.O) IF(WP.0 RFGTI .0) IF(WM.( WM W2 0.VI + V2 WW = WI 4+ W2 PPX .5*( UP VI 0.L1.PPX $ PPZ + PPT GO TO 500 IF( PSUM.UP.GT.
0.EQ.oPF7Z)'PFT UIs .( i.0.0PET) Vi*PFX*( 1.NE.0 ) GO TO 600 DIFRA(I) .0PFT) V t.0PFX)#PFT V2( 1.0OPF Z) W2 *P FX' IF Z WI'.PFT) WI#PI'X#PFZ WI'*ix( .*( PPX + PPZ )/PSUM GO TO 30 600 OIFRA(I) .PFT u i.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$( t.PrT )/PSUM DIFTA(I) . ronTRAN A FILE: CON4VERSATIONAL MONITOR SYSTEM DFPRG AVM/SI' FOTRAN DI F00560 DIF00570 PACE 002 PG 0 OMT GO a 0IFT'156 TO 999 00 PFX PF Z *0.o VELX( 1) VCL~X(7) VELX(3) VCLX(4 ) PFX i'rz PFT   U1/(UI'U2) Vt/(VI14V2) Wl/(W14W2) r DoIooseo 01F00590 DirooGoo.X(I)'DELR*( PPZ + PPT )0 'PSLJM DIFZA(1) m VELZ(I)*DELz#( rrX +.PFX*PFT V2* I'F X'PF T V2.o 'FX'( 1.FILE: DIFPRC'.( .0 (Vl*V2)..0PFX)'PFZ Wt'(I.V(LT(I).0 Pr T Ir (UI+U2).0PFX)o( t.0 oirZA(l) .0 Ir (W1#W2).0 PPX .VEU.09'FX)#(t1.0 oiETA(I) .0PFX)#(t.0PET) Vl*( 1. Z(2) vrI.NIE.0PFT. oi rooG to Dl F00G20 D1F00630 01F00640 Ut 'PFZ .0 30 CONTINUE RFGRI RFGR RFCTI  DIFO 10 tO 011:'01020 DIFO 1030 01 FOt040 DO 1050s DI FOtOGO DIFO 1070 01000 DIFO 01 FOtO0J0 OIFOi tOO DIFRA(5) v DIFfIA(t) aDIFZA(2) + DIFRA(6) + DIFRA(2) + OIFZA(3) DIFPA(7) + OIFRA(3) + DIFZA(6) + + + OIFRA(fl) OlfRA(4) + DIFZA(7) r % I .0.0PFZ)*PFT U20( 1.VELX( I)/I)ELR PPZ .0PFz) W2 &1EX'&( 1.0PFX)*PFZ DO 30 11.0.0.VELT(I)/OELTH PSUM a PPX + PVZ + PPT IE( PSUM.0.OPFX)O( I.NC.O.) U2 4PF Z PET 1J2'( 1.0PFX)*PrET V2*( 1.IPFZ)s( 1.0PFT) U I PFZ (I 1.0PET) Vt .DELTH.0PFZ)'$( .0 0.0PFZ) W20( 1.0PFT) U2*PFZ*( 1.( 1.0PFZ) DIFO00650 oDI roofGo oir00670 nir006s0 mrof0069 ot rooioo DIF00710o VEIX(S) VELX(6) VIELX(7) VELX(0) VEL7( I) V El.VELZ(I)/DELZ PPT .
511S ORRY) 200 STOP ENO DIFOi IfO DIFOI 120 DIFOI 130 oDIFO140 DIFO 150 DIFOI 160 DIFO 170 DIOt t0100oo oIro 190 DIFO1200 DIF01220 I I .2) G FOIMAT( III .RFGR.2X.RrGT I.I.2) 7 FORMAT( ItII . 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 .7) 5 FORMAT(9F5.6F5.RFGT.OMT I.20X.OMT GO TO 200 500 WRITE(G.
C .I KN a (X(I() X(K) RE TURN END  X(K41)*COK))/O(K) VAR09Z350 VAflo9:160 VA1109370 VA11093001 VAI.O(N).0t4O01 X( I ) I~) 13( ) 0(i1) NM I *N I 00) 2 1I.X(N).NMI .P(400).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.C(N).D.1O9:190 VA1109"IO() VAn0094 t VA11O0d1O '4A110943?0 VAIIW14 40 VAfl0O450 VAI*0946) VAI?94 70) VAR0O9400 VAR094 90 VAIR095C VAflfO!'j 10 VA110IJ520 .0.XN) DIMENSION A(N) .o(N).SUarIouT INE TDM(A .
]K . 1) GO TO 400 IF(K.KBP2 " (3.CR..UG.K) a LUGDX... iB10 .E230 .E1O 0005 FP" * IF(.L..0 OOC COI~'T2]'UE C VAR P182 vLR a210 VAR E220 VAR.2D LUGRDX/(AUGOX +~DG'RD)' . DA'' G.FFX ( ..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.C "hT.)K2NC " (Ki)'NWPC .K) * 0.I.K) z 0.I.) G0 TOrOOiC D'.VGRD2/(.PM.G  V RIS150 VAR.ItD. I O50 VI8090 K1 K12 2 KEP CONSTA'NT CLLCULATION C CRO5SFLD'.JSP2 DO O40000 DO 4000 K. I vtR I5S0 rooOT 00G2 UC 'S(U(]K) ) VLRi 1210 vAR1E230 VAR E20 U(3KXX) ) LUGc .E0. 1CONV.LT.P2) GO TO 40002 (U(iK)*U(KX))) LUG = LS(O.3.F2(J.P.DFFUSI'O =.TIONS SECT ION AD !ST ]F(ICLL].NPC o1 ]F(K.KZ2))) .K) • AVG.5(V( K)*:(.LE0.EO. C C C C U AND v LDE VELOCITY EDU.G.LT.S(0.FX(3.KS 2.K2NC JKZ2 .EO.1) GO T0 49998 SOLUTiON FOR SIE.K t KXX a jK .0.S( 4 .L!S( (]KZZ) ) ..UG. D2).xx'.E0.C .:X GO TO 40000 "0OO D.O.0 D'rF2( I .S . DNF.
E003 DO E0100O ) K GO TO50205 222. ( '(K ULH C CON'..KSPI I KL I lK z lK 1A 0.. UD( 3K) UL UR UT Uo PK) UD( 3KP) U0( 3KMS) DCR DCL D!FFCD(IKL) D) F CO( KL.1.I ) I KIPC # (KI)'tN'PCL VAR16D VAR 11510 VARi18520 !PK a 31.5 )r0. .E.. KS I KLS a3)K I KP a IK  NP 2 I N1PC VAR 185501 YLR 15560 VAR 1C570 VARias590 VAR 86201 VARi1Es30.5 PC) '). 054 (K.'O('.2)GO a TO 29990VALR iS.C RD20 3KS0 S3G'~ to VAR 8340: VARIS5360' RDZDRD2 VAR i !:0.S(UftH)(UCUQ)  VL21S660.1 SRKSrI a * N BUOJY a LP0E7.I DIFFCD(3KL2) D!~FFCD(IKL3) *0(!.]BR ( '.0.0.5 UODUMKS) 16D(1PK) * .ET/TADD3 VAR IS420 VLO IS440 3F~lDw.I) K2WC' + II)K NC *K2NC "00Z0100 K=2. 3F(3~.PKKNUO(IPK) ).P( TJ"'.R2CLI. a uJD(2K) )0. VLRIE6t0 V4RI15650 VAR 15690 ULWM'(uLLIC) ( IR'(L'C4UV) VARIE *7 DO.
'N(tX) + 06Md~7A 1 Et a 6Nr NZ At 2 SW SWA r r ). (W (4I) 09O6t~u7A CA ~g0 *0 H IL~ Iun S:)o ( (dN I)an +(>(!M)Onl 'S0 06SLIA OEOG bVA 0L6O t'7A 0CS8~ 100 VA )O (dm)OoA1 0669'.) .=OStd7A 0506LdV  ~ ~ ~ ~ Hm )CC7r~! Sall*~ ZN07J ±. 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 )*.+fl)IdM Y(Qd.o.OSZ6 M7A OdI'N Al dN O7I~t 17A OdM ( A) ONZ Nt + 1IOZN4(Vt) * + 3NZ)(11) + t bi > >4 E A c TA ..7A 0060tMIA 0L6Sgv7A 0968LVA I 3a CA~ SimI) 0L96 t'7A O9S63tM7A j0 0 VA d061Nt) AoflsaNx0As + Od.(?nOn) i13( fl2f) unc.8~ VA 0899tMVA (oaQN 3NC647.vt Z1t0 7 Z(J~Si0(o(nsn)(a)a (1n.uQ + + t (Nt)0ftl x = XIt)f ( oo.% At t x NdE ONCA + lA b + A t + OZ~N(Lt) OdAN 00gla aVA 06!..(.( (nf1l)(Himlss7 (In3n)(uLM)S9V ) ZNQ..o2m.(onin) M0'a(On.onH.(iim1 (.t) 2 td9N'Zt oo0to oa 3flN tLNOD s0?_O3 M 6 dVA + NAcns + ZntO * (Nt)d(dXt)d ).)10 6666a 01 0 MfN i NGO 07 6 "VA ZNOO NCO ZOM( + (Xt)0M £ ) O~t6tVA (~ ~ .Cw'))'A.
NWPC DCT 3 DIFFCO(IKL+1) DC.1SP J a (K2)]IeR * 1 .KSPI ]K = 1 + (11)K2NC + (KI)NWPC 1X a (12).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 .IMKS IKt.AeS(WRB)) .$ ]K .( DCT + DCS ) C1(J) a 0.]K .KBR + K 81100 U(IK) a X(IX) C DO 82000 K=2.ROT * 0.5 AI(J) a 0.X.SIGMhARDZDODCB E1(J) = RDT * 0.RDX.ULT * ABS(ULT) 81(J) .5RDX( URT 1 + SIGMARDXD.5( U(IK) + U(]KP) ) ULT C a 0. C 00 l8100 I=2.5.5( U()]KS) * U(IMKS+NWPC) ) Ai(J) = 0.5RDZ*( VRT * ASS(wRT) .8e1.KER D00 82000 12.KBR DO 62100 1=2.]K + K2NC IKP a 1MKS 1K + NWPC ]K K K2NC NWPC 1KIAS = C DCR .VRE + ABS(WRB) 1 * SIGMA*RDZD.DIFFCO(IKL) DCL a D0FFCD(IKL+2) URT x 0.'5 WRS (W(1KMS) * W(IPKNWPC))O.X.'RDXDDCR D01(J) . VAR19310 VAR19320 VAR19330 VLR193AO VAR19350 VAR19360 VAR1S370 VARie380 IKL a C 1 + (1)'K2NCL + (K1)NWPCL IPK .5*RD2( WRT .IKBR) CALL TDM(A 1.( DCL * DCR ) C1(J) z 0.5RD02( WRe .ABS(VRT) ) .SIGIArRDZD3DCT 01(J) a U(IK)ROT 81000 CONTINUE C ..SIGt.0.5*RDX( ULT . = DIFFCD(IKL+3) WRT 3 (W(3K) + W(iPK))O.K2NC .DI.W(1K)RDT 82000 CONTINUE C CALL TDM(A .16PI 1K a 1 * (]i)K2NC + (K1)NWPC IX x (K2)ISR + I 1 E2100 W(]K) * x(Ix) 00 823000 K2.( URT EAS(URT) ) .IBR DO 81100 K2.C1 .SIGMARDXD*DCL ACBS(URT) .1 ]K £ + (1])K2NC + (Ki)NWPC ) VAR 19260 VAR 19270 VAR19 280 vLRlg2O VAR 19290 VAR19300.ABS(ULT) ) .C1.KBR) C DO 82100 Ka2.
LG!1CE)PA HIM.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 .Htin 0~Q0~*/'.v4)tS ~m * io1..)ssv .v%4!)iS +.~tSC(Hifl)ss b3(3 ).'%)SS'7 + HSA . ). 3)3QO. )'ZQdS50 )Zc!S*  a ' a  (HS.3 00 000O~d7A "L661 aVA (AI  (Ad 1) d !OSGOZi VA OG61 !VA 03661 SVA b33aXac~tjS ( 100 (Hll)SV H~ln   C(Hzn)sev  Ya   ~ A +Laa Hfl )xCi~s~o = x (r)La ()9 t ()1 2 (H.Q ( (S'feXtiM 094=J7A OSZO0t7A (>I).' ).fl.0xCd.L 0.3nwNtJ.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 ). z (r 01)~S~7~).(1) L >1 * (~~t A4d! zz IN I s)Zt! + Afl9nszAofl3 (SCOP 001 0OZ17A 060ZZ2A OSOOZZ17~JA 1 OdIA'Na)) *  AC470103t1)1E 0*0 a NAC(3 a9AzsA 007 OC Idgl*Zal OOOVS 00 OLOOZ!7A 0900ZM7A * ()t(z>q) :)NN41) + I a as9tz:t 001C3 + Cxt)y 2 (x)Xfl( ( 00t ES A4t cc tdaA'Zx4 001.~) Al NP 1=I000ca 00 .40 C0~7A O CO0d7A NAoC1s (A(I)a .mNtAz + + ONA(lt.(HL')S.
HDX RRC a I.CALL TDM(A .SIE(IMKS) SIET .2.B.X.1CALI.0/RC OR = CYL0.NE. 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 .S) SIECO . VAR20420.LE.DIFFCC(IKL) DCT .U(JMKS) vW = W(IKMS) AUC .DIFFCO(1KL*i) DCL a D!FFCO(1KL+2) DCB * D]FFCO(JKL*3) ]PK )KP a a + K2NC * NWPC = 1K .K2NC ]K IK 7]KS IKMS = IK .0P.SIE(IKP) S]EB * SIE(]KW.KSR IK IX a I + (I1)*K2NC + (Ki)*NWPC + K .0.O.5(TS(IK)+TS(IPK)) .1SP1 00 84100 K2.EO.I) GO TO 55000 DCR .12 J .SIE(IPK) S*EL .Ci.1 (12)(KER1) 84100 W(JK) a X(lX) C 89999 CONTINUE IF(ICONV.1) GO TO 2040 ISEO = ISEO + 1 IF(ISEO.25*RRC CYRX = CYLO0.K2 DO 60000 13a1.NWPC CFR = CF(1PK) CFL . IBR(K2) + I ' (1i)K2NC IK = i + (K1)NWPC IKL a 1 + (11)K2NCL + (K1)NWPCL RC a FLOAT(I1)bDX .CF(]MKS) UC c U(IK) WC = W(!K) UL .DLeS(UL) ALC = DAES(WC) 1wS * DASS(WE) = SIE(1K) S!EC SIER .IKBR) DO 84100 l=2.EO.5(TS(]K)+TS(lMKS)) VAR203G0 VAR20370 VAR20380 VAR20390 VAR20400 VAR20410 .S]EO(]K) TSRA TSLA .DASS(UC) AUL .CF(UK) IF (CFC.D1.NSEO) 33333 CONTINUE 15E0 a 0 DO 60000 K=Ki.5RRC*RDX CFC .
I vo 0009 0:9>4 I )c t a.(r'. C0 r~t (r)LC 00~ttIdVA 060k ~wVA 0OLt dVA 00009 01 13t5.+c a 21J10 to.x)j! 2(AI1)3rS oo0t 9 06CI ~1VA' 09ZI dVA 0LI.NOD (L03e.Q ~ 000 fl co(3 S6 .OSrlth7A OV7 t!VA t. N *:dM o zi Z)q4s)4 coo 0. t7) .At~VA OSZI~7A clt'L fa ZAVtAZ'4 0=0i9 00 0=O9 00 I .(N t0 *090~bVA 0O0I bVA 0O'0t MVA OSOI aVA 0OP0tZSVA (Q QtlY.± 4:7) 21J3 1u L7S1L'q'l OS 60ZMVA (Slx () I s I)SO I ((diE)l+ m1)l)'c = vzs I VIL .C*~ t )4* ."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.~t J '~1f ((t0AX'1ZJANO ) C dt0* ) * N ) S 4~ I . 06C Zd7IA ZltrEzf 00 OOOL CO 0SCtLb7A NOI0t1D3JCZ NE d33MS CNZ 0L~tZZ7A LCECE OZ :tZ7A 1 05 3 3lNTI. (nN*31.L.OlI)03(r v+n ..1! Al z(IJ)L3 (NOfl'Ol V(ISJX017) Ot)t0dVA 0001ZlA N7 dlxiol ((L1)x~JNG4(IV~~sNQ)ts0 . t :) a . 10 + ( 0S669 01 00 (V03VIJfl S0.
0 GO TO 70000 65500 C2(j) E82(J) .5(rS(lK)+TS(lKMS)) C GAT DIFFT DIFFB C C * RDTSIE(]K) .5. .k)+DNFFZ(1.IF C DCR OCT DCL DC8 C ( CFC.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.NE.11.5(TS(]K)+TS(3KP)) TSeL .C2.K.I) )XMX GAMTTSTA VAR21620 VAR21630 VAR21640 VAR21650 VAR216GO0 VAR21670 VAR21680 VAR21690 VAR2I700 VAR217 tO VAR21720 VA.TTSPLL ..0.I DI(J) a TGA.RDZ C2(J) = 0.0.LE..0 E2(J) .5 0.5"(ws*Aw )'RD2 .DIFFCO(IKL+3) 1PK a 1K + K2NC IKP a !K + NWPC IMKS a 1K .5(WCWC).0 C2(J) = 0.(A2.21730 VAR21740 A2(J) = B2(J) = VAR21750 VAR21760 VAR21770 VAR21780 VAR21790 VAR2800 VAR21810 VAR21820 VAR21830 VAR21840 VAR21850 VAR21860 VAR21870 VAR21880 VAR21890 VAR21900 VAR2190 VAR21920 VAR21030 VAR2190 VAR21950 VAR21960 VAZR21970 VAR21980 VAIR21990 VAR22000 .C2(J)SIE(IKP) C2(J) = 0.NWPC C CFT = CF(IKP) CFB .M * * IF(TS(]K).1) GO TO 65500 D1(J) * 01(J) .X GtA.CO(IK)0..B2.RDZ .K+1))X.(WC*AwCWS*iWB).K2 Ix = KS.NNOPT.ND.X.DIFFEDCRD2D RDT * (DIFFTDCT+D)FFS*DC6)*A02D .K)*DNFF2( .(12) + K .NU) GAMT * RPRAN .0 DI(J) D(J) + DIFFTDCT*RDZ GO 70 70000 65000 01( ) = SIE(]K) L2(j) a 0.1.0 69950 IF(CFT.IKBR) C DO 71000 ]=m1. 1 * 0.5(DNFFZ(I.0.DIFFTDCTRDZD C2(J) a 0.K2NC IKWS VAR21460 VAR21470 VAR21480 VA021490 VAR21500 VAR215 O VAR21520 VAR2rS30 VLR21540 VAR21550 VAR21560 = ]K .1) GO TO e9950 Dl(J) = 01() .12 DO 71000 KKI.01.EO.DIFFCO(IKL) = D]IFFCO(IKL+1) = DIFFCO(]KL+2) .EO.E0.A2(J)SIE(3KMS) A2(J) • 0.i) GO TO 70000 IF(CFT.5(DNFF2(3.OlFFTDCTRDZD IF(CFB EO.0 70000 CONTINUE C CALL TDM.1) GO TO 65000 .
K2 LWPC=LWPC+NWPC AT TIMEN ) . 12 KK=KK4K2NC KKL VAR22010 VAR22020 VAR22030 VAR22040 VAR2 2050 VAR22060 VAR?22070 VAR22080 VAR220s0O VAR22100 VAR22 110 VAR22120 VAR22130 VAR?22140 VAR22150 vLR22160 VAR22170 VAR22180 VAR22190 VAR22200 VAR222A0 VAR22220 VAR22230 VtR22240 VAR22250 VAR22260 VAR22270 VAR22280 VAR22290 VAR22300 VAR22310 VAR22320 VAR22330 VAR22340 2109 VAR22350 VAR22360 V R22370 VAR22380 VAR22390 VAR22A00 VAR22410 = KKL + K2NCL VAR22420 VAR22430 VLR2240 VAR22450 LWPCL = LWPC=I 1 VAR22460 SIE( 1)=S [(KMS) U(1)=U(!KMS) w( ) (IKVS) 0(l)TO(IKMS) TS(i)=TS(IKMS) SIE(]K S)=SIE(KK4+ ) U(KtMS)U(KK+1) W(IKt.S)=W(KK+) TO(KMS)=TO(KK+1) VAR22470 VAR22480 VAR22490 VLR22500 VAR22510 VAR 22520 VAR22530 VAR225ZO VAR22550 .E0.NWPC DO 2. 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.11.IK r 1 + (II)*K2NC + (K1).5'( SIE(IK) + STOR(I.O) 00 33335 I11i.SIEO(IK) KIRK a i GO TO 33334 33337 CONTINUE DO 33338 1=1112 DO 33338 K=KI.K2 1K 1 + (]i)'K2NC + (KI)*NWPC 33338 SIE(K) = 0.K) = SIE(IK) 333335 SIE(IK) . TRANSFERS VELOCITIES TO STORAGE ARRAY( 2040 KI1 K2xKBP2 LWPC=I .K2 IK v 1 + (1I1)K2NC + (K1)*NWPC STOR(I.1) GO TO 33333 GO TO 33339 IF(ISEQ.09 KzK1.12 DO 33335 K=KI.K) ) KIRK a 0 33329 CONTINUE C C NOTE.NE.NwPC 71000 SIE(]K) = X(IX) C IF(KIRK.
is for calculating the natural convection flow field in a Cartesian closed compartment.6 and Fig. . 'inputd HUH' and 'inputd DM'. 'inputd 3' and 'inputd 7'.3(B). results are in Fig. Fig. 4. 'inputn dat'. 5. The output The files. 5.APPENDIX D TYPICAL INPUT FILES The file. is the input to test the stability of the ADI scheme for momentum equation.1 and Fig.7.8. 'input mom'. 5. 5. The output results are in Fig. are for comparing Huh's And De Vahl Davis and Mallinson's correction formulas in the problem geometry. The file. 5. Fig.2. The files. are for calculating the numerical solution by donor cell scheme in the problem geometry.
01 0 1.0 4.0 2.0 0.0 1.0 0.75 10.0 0.0 0 0 1.0 1.7 0.0 0.0 0.5 6.09375 0.0 0.0 0.0 0.0 0.0 1.0 60.0 0.0 0.0 60.0 1.0 0.045 0.0 10.0 1.0 0 2 1 3 O0 2 2 O O ADI 10.0 2 l.5 0.0 0.t20 1.120 1.4 1.0 1.4 5 1 1.4 0.0 I 3 I 16G 1.0 0.4 0.0 1.0 0.0 4 0.0 .44 0.0 0.0 10.2 10.0 1.0 0.01 0.0 0.rTh FILE: INPUT MOM VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 4 4 O STABILITY TEST FOR MOMENTUM 2.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0 60.9 0.0 0.0 0.0 1.0 2 5 10.0 0 0.0 10.0 0.013 0.0 1.
0 1.0 6 6 50.751.+20 60.0 0.0 60.0 0.0 677.0 0.0 0.20 0.5 1.20 030 1.0 1.0 1.0 8 2 I 3 1 3 0 3 0 1 1.045 0.20 2 1.7 1.FILE: INPUTN DAT A VM/SP CONVERSATIONAL MONITOR SYSTEM 'PAGE 001 5 10 0 1 0.2 1.67 1 1I 2.0 68.0 32.90 0.0 1.20 1.0 0 1.5 0.0 0.0 0.0 0.0 4.01 1. 20 1.0 1.01 0.0 I 43 1.0 2.20 II I 1.0 1.21795 1.0 1.0 NATURAL CONVECTION IN A CARTESIAN CLOSED COMPARTMENT 1.5 5 1.0 0.0 0.0 0 0 2 1.0 1.0 0.0 0.0 10.4 0 0 0 0 O 0 0 6 2 0.20 1.44 0.0 .0 1.
0 4 I 36 1.0 0.511 1 1.0 1.061 1 0.0 1.07107 7.0 1.911 3 4 1.01 0.000 0.5 0.0 1.0 0.9 0.0 1.000 0 0 O O I.000 1.063 0.0 9 10.CASE 3 2.0 7.0 0.997 7 1.0 12 11 2 121 t 5.0 7.O.002 0.01 0 0.0 1.+20 I 0.0 9.0 1.OGI 2.204 0.0 0.0 10.063 ( ).0 1.0 0.014 0.0 0.0 9.940 1.0 0.0 0.0 1.0 2 4 2 4 1.07107 111 3 1 1 2 5.0 0.0 2 1.000024 0.0 0.70711 1.0 0.000 10 1.0 0.0 9.0 0.05.0 9.0 1 .0 0.5 1.09375 0.0 1.07107 0.014 ).0 0.0 0.974 5 1.0 5.0 0.47.5 1.000 9 1.207 0.0 7.249 3 1.003 7 1.0 1.07107 0.FILE: INPUTD 3 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 10 10 0 0.0 0 2 5 3 0 7 6 1.3 1 I 1.0 0.204 0.0 GO.0 0.000 10 1.497 2.991 6 1.999 1.0 7.001 0 1.0 0.0 0.07107 1.751 2 1.0 0.026 5 1. 020 2.013 2.II I 11 2 2 1.0 0.000 0 0 0 O .0 0.70711 0.207 0.0 1.0 0.457 1.0 0.07107 12 12 Ill 2 5.0 1.0 0.0 0.0 0.0 1.0 7.0 0.457 1.0 1.0 9.0 0.0 5.0 0.0 60.02 1.002 0.000 0.07107 0.045 1.0 1.0 5.0 1.009 G 1.0 0.0 1.0 0.497 0.409 1.44 4.O 2 1.75 0.0 1.0 2 1t 1 113 5.0 8 10.000024 60.0 0 0 NUMERICAL blFFUSION .948 1.0 9.0 0.0 .0 0.0 0.0 0.009 4 1.
0 7.0 0.0 0.0 1.5 1.0 3 0.0 0.013 4 1.0 0.07107 711 8 8 2 5.0 1.0 0.47.0 0.0 5.071107 0.001 0.0 0.0 0.0 0.457 2.0 0.09375 .0 NUMER ICAL DIFFUSION .0 60.0 0.0 10.0 0.001 0.0 2 7 8 5.0 2 9.0 0 0 2 2 1 0.0 7.0 5.626 1.0 1.0 0.0 2 2 0.0 2 0.01 2.0 4 2 4 1.00.0 0.014 0.0 7.0 1.0 0.0 5.07107 0.0 0.0 1.0 .420 1.0 1.07 07 0.457 2.0 0.0.0 1.013 2.0 1.333327 1.0 0.3 1 1 1.9 0.002 5 1.0 4 I 36 1.4 8 2 1 3 0 0.0 1.633 1.0 1 9.0 0.44 4.07107 1 1 713 2 5.003 0.911 9.75 2.000 0 0 0 1.374 1.0 9.014 0.0 0.0 1.0 0.0 7.045 1.FILE: INPUTO 7 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 6 0.0 0.4 1.000 6 1.907 3 1.0 1.000 6 1.0 0.0 0.0 10.000 0 O 0 0 0 7 1 2 7 2 5.990 5 1.0 60.CASE 7 0.0 0.143 0.0 7.0 I 1.0 1.167 1.0 0.07107 I 113 7 2 1.0 0.0 1.333327 60.07107 7.0 1.5 1.0 1.+20 2.003 17051 1.17051 0.0 1.01 0 0.143 0.009 1.0 1.0 0.0 0.00.0 0.0 0.0 3.633 1.01 0.0 4 1.
0 9?LL '0 G'.1 1' 0*0 0 *1 D99C00 0'1 0IO' 0*0 0*0 0*0 0'0 0 *I0*0 9c 1 t.L Z01 'L L'09C91C P09c91''? 1'Q9C9p 0'19C91' 1 iLO'O 1'LO*0 GL00O LLO'0 £00.09 W '09 0. 6IP'0 O1'1 991'0 LOCO0 9LV'O 060*0 Cc£00 910'0 1 0 C Iz 0 0 0*1 OL+*t Ylz OZ4*1 zlo' IfIAH .NOIS(fJJIG IV:)IUIW(IN too 39vd IVNOIIVSU3ANIX) dS/W~A v VIIi III)II OIMINI O~N :311A II .0 0 *I O'l 0' *1II1 01l 0*1 0 1 0'L 01 O'1 0.0 0.0 CGO.SAS 3~~~)Vd kOIINOW 0 0'0 0*0 0.z 10'o 0 1 0'l GL'O 1O1 '0 I 0*0 01l 0'l L910* 199*0 099.u *0 Ul' I 6P I 0011* I 0. 1'0 0. *0 PI 9..1bl b 690*6 £00*0 LC9 0 90c.0 09G"0 9Cs£0 661''0 6sv1'.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.*I riI 00 0*0 0*0 0i'00 0 10*0 6 v 0*0 0'0 top' . 0 LG0 .0 GLO10 ZU10 SL9*0 109.0 01l 0'00' C0'I C.09 0. 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 5.20 911 1.0 5.20 1 1.0 2 25 5.0 I 1 5.20 2 1.0 2 25 25 5.0 8.6603 0.20 10 1.20 5.20 9 1 1.6603 0.20 1011 1.0 .0 0 2 1.0 0.20 2 1.0 0.6603 0.0 0.20 I13 1.FILE: INPUTO IIU4U VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 2 0.0 26 26 5.20 913 1.
4 4 1 36 1.0 0.0 0 2 1 3 0 7 6 I 0.GOG 2 1.005 0..0 1.0 1.0 2 4 2 4 1.0 0.30i9 1 0.4GI6 0.0 60.0 0.0 0.000 0.0 0.002 2.403604 2.949 0.0 0.0 2 2 2 0.0 1.636 0.02 2.0 1.767 0. 1.009 0.003 0.0.076 0.0 0.260 0.0 1.0 1.093 0.0 5 1..463G04 2.133 10.0 0.0 1.0 1.120 2.0 1.0 0.3 1.700 .0411 0.02..920 0.0 1.0 60.9 0.0 0.030 0.0 0.20 1.0 1.463604 2.OM 0.0 0.0 0.0 1.0 0.566 0.75 1.903 0.072 ' II 12 13 14 15 16 17 10 19 20 21 22 23 24 0 0 0 I '.034 0.0 0.0036G0 0.099 0.705 0.01 0.420 0.365 9.115 0.0 0.2406 1.0 1.223 3 1.0 0. 4G1IG04 7.5 1.0 0.0 1.002 0.360 0.0 5.002 0.rILE: INPUIO OM A VM/SP CONIVEIlSATIONAL MONI(fro SYSTEM PAGE 001 24 f( 0 I 0.156 0.463704 9.0 1.0 1.0 0.33G 6 1.44 4.0 1.011" 0.0 0.0 1.5 1.0 1.567 9.0 0.0 1.030 0.0 5.951 0.0 0 0 NUMEtICAL IrFFUSION .0 0.0 0.0 1.0 0.003 0.0 3 10 1.09375 0.0 1.969 0.0 1.021 0.460 4 1.013 2.144 1.932 1.051 0.003 0.0 0.003660 60.4 0.016 0.526 3 4 5 G 2.174 1.044 9.496 0.0267 0.OGI 0.0 0.302 0.02.0 1.522 0.242 7 1.029 0.0 0.0 0.007 0.0 0.0 1.045 1.0 0.197 0.01 0 0.0 I 1.0 1.0 0.463604 2.5 1.0 1.0 0.246 0.96G 0.0 1.026 0.
0 5.20 1.: INP.SILr.0 I 5.0 2 5.10 OM A VM/SP CONVERSATIONAL MONITOnr SYSTEM PAGE 002 2 0.20 9 1 1.20 113 1.0 26 2 1.0 0.0 5.0 . GG03 0.20 911 I. & .20 913 1.0 I*  .0 26 5.6603 0.20 1011 1.20 I 2 1. 6603 0.20 5.0 0.0 0.0 2 5.0 0 29 25 I 25 2 1.
This action might not be possible to undo. Are you sure you want to continue?