This action might not be possible to undo. Are you sure you want to continue?
Kang Yul Huh and Michael W. Golay Energy Laboratory Report No. MITEL83011 August 1983
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS
by
Kang Yul and
Huh
Michael W. Golav
Energy Laboratory and Department of Nuclear Engineering Massachusetts Institute of Technology Cambridge, MA 02139
MITEL83011
August
1983
Sponsored by: Boston Edison Co. Duke Power Co. Northeast Utilities Service Corp. Public Service Electric and Gas Co. of New Jersey
2
ABSTRACT
TREATMENT OF PHYSICAL AND NUMERICAL DIFFUSION IN FLUID DYNAMIC SIMULATIONS by Kang Y. Huh and Michael W. Golay
A computer code is developed to predict the behavior of the hydrogen gas in the containment after a lossofThe conservation equations for the four coolant accident. components, i.e., air, hydrogen, steam and water, are set up and solved numerically by decoupling the continuity and momentum equations from the energy, mass diffusion and turThe homogeneous mixture form is used bulence equations. for the momentum and energy equations and the steam and liquid droplets are assumed to be in the saturation state. There are two diffusion processes, molecular and turMolebulent, which should be modelled in different ways. cular diffusion is modelled by Wilke's formula for the multicomponent gas diffusion, where the diffusion constants are dependent on the relative concentrations. Turbulent diffusion is basically modelled by the ke model with modifications for low Reynolds number flow effects. Numerical diffusion is eliminated by a corrective scheme which is based on accurate prediction of crossflow diffusion. The corrective scheme in a fully explicit treatment is both conservative and stable, therefore can be used in long transient calculations. The corrective scheme allows relatively large mesh sizes without introducing the false diffusion and the time step size of the same order of magnitude as the Courant limit may be used.
A
3
ACKNOWLEDGEMENTS
This research was conducted under the sponsorship of Boston Edison Co., Duke Power Co., Northeast Utilities
Service Corp. and Public Service Electric and Gas Co. of New Jersey. support. The authors gratefully acknowledge this thank all those who assisted them
The authors
during this research especially Dr. Vincent P. Manno, Mr. B.K. Riggs, Ms. Rachel Morton and Prof. Andrei Schor. work was expertly typed by Mr. Ted Brand and Ms. Hakala. Eva The
2.1 Continuity and Mass Diffusion Equations 2.1 Diffusion Convection 4.4.4.1 Molecular Diffusion 3.3.3 Basic Assumptions and Limitations 2.4.2 ke Model 3.5 SMAC Scheme and Compressibility CHAPTER 3: DIFFUSION MODELLING 3.3 Truncation Error Diffusion in a Two Dimensional Problem 4.2.2.1 Problem Statement 2.2 Crossflow Diffusion 4.2.3.4.4 ADI Schemes Conservation 4.2 4.2 Physical Constraint 4.3.1 4.4.3.1.2.1 Mesh Point Implementation Mesh Interface Implementation 4.2.2 . Implicit and 4.1 Derivation of Turbulence Equations 3.4 4.1 4.3 Discussion of Skew Differencing and Corrective Schemes Raithby's Scheme 4.1.3.2 Turbuletit Diffusion 3.1 Skew Differencing by Huh 4.4 Solution Scheme 2.3 Tensor Viscosity Method Corrective Scheme by Huh 4.4.2 Truncation Error Diffusion 4.3 Low Reynolds Number Flow CHAPTER 4: NUMERICAL DIFFUSION Truncation Error Diffusion 4.3.1 Truncation Error Diffusion in a One Dimensional Problem Another Approach for a One Dimensional 4.1 Assumptions 2.2 Energy Equation 2.1.2 Limitations 2.2 Comparison of Explicit.3.2 Governing Equations 2.3.4TABLE 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.
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
17 Comparison of the analytic solution. . skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure . . . ..3(B) .3(B) .430 . .15 . Donor cell solutions with various diffusion corrections for a pure convection problem with 8 = 450 . 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. .11 Comparison of the true solution and solutions by donor cell scheme. 144 . (D+DDM). 136 5. .. . by De Vahl Davis and Mallinson's formula along the line BD in Fig. . 137 5. . . 5. ..convection problem with 6 = 450 Comparison of the true solution and solutions by donor cell scheme. .. 5. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with 6 = 63. . . .. .43* . .8 . .14 140 5.9 5. True solution and donor cell solutions with and without diffusion correction for a pure convection .. .10 Comparison of the numerical solution with donor cell differencing of convection term and analytic solution with increased diffusion constant. problem with 6 = 63. . .3(B) . Comparison of the analytic solution and numerical solutions by donor cell scheme and skew differencing scheme by Huh for 6x6 meshes along the line . 5... .. . . 135 5. . CA in Fig. . . ..16 143 5.12 . . 142 5. 5. numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig.13 139 5. . .
.. . . 152 5. .. .19 . . 5.. . . . .. .. Simple recirculating flow field for test of the implementation strategies of the corrective scheme ..19 . 146 5.. ... . .22 150 5.10 5.. . Diffusion constant corrections at each interface for the mesh point and mesh interface implementations of the corrective scheme for recirculating flow field given in Fig.21 S.20 .. . . .3(B) . . 5. . . .18 Comparison of the analytic solution. 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.. .. 153 . . . .19 . . . .. numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig.. . ...147 5.. .. 151 5... . .. .25 Diffusion constant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment . . 148 5.. 145 5. . . . ... Recirculating flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme Detailed flow field for calculation of the steam concentration distribution in the two dimensional containment to test the mesh point and riesh interface implementations of the corrective scheme Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .. . Comparison of the true solution..24 .... . 5.23 ..
.0 sec and the Courant limit of 1.. . .30 for At = 1.29 Steam concentration distribution after one time step. 5. . .cx T where 4 is given by.. 156 5. . ..3 sec. .. with donor cell differencing of convection term without any diffusion correction .5 sec and the Courant limit of 1.. . . . Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. . .0 sec . Steam concentration distribution after one time step. . . . .. 5. 5. . 163 A. . 157 159 5. . . .. Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig. .3 sec.26 Diffusion constant corrections of the mesh interface implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment . . 0. . . Problem geometry for comparison of the explicit and ADI schemes . .33 Comparison of the ADI solutions for the profile of 4 at t = 10.0 sec . . .. .27 . different time step sizes Boundary conditions in terms of 4 for the problem in Fig.28 . 1 178 . ..30 5.32 161 5. Steam concentration distribution after one time step. . ..31 1i60 5. . 0. . . .3 sec. with mesh point implementation of the corrective scheme . . .. .1I 5.. 5.30 with ..0 sec along the line AB in Fig. 0. 154 5.3(B) = e..30 for At = 0. . . . with mesh interface implementation of the . . . 155 5. . . .. corrective scheme ..
1 List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research. . . . . 78 4. . 1981 . . . . . 2D. . . . . . . .12 LIST OF TABLES Number 4. . Page . . . .114 . donor cell differencing of the convection term . . . . . . .2 . Comparison of the Von Neumann analysis and numerical experiments for the stability condition in terms of the cell Reynolds number in an explicit. .
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 .13 NOMENCLATURE Letter A c c Definition Area Specific heat at constant pressure Specific heat at constant volume Courant numbers in x and y directions (1) Dimension (2) Diffusion constant (3) Divergence (4) Discriminant of a quadratic equation Binary mixture diffusion constant between air and hydrogen Binary mixture diffusion constant between air and steam Binary mixture diffusion constant between hydrogen and steam Crossflow diffusion constants in x.d xy e f Gr h I Im k g M Nu P Pe p y Az Pr Q Prandtl number Heat source .D y./Imaginary (1) Turbulent kinetic energy (2) Thermal conductivity (3) Ratio of specific heats.D d . cp/c Length Mass Molecular weight Nusselt number Peclet number Effective Peclet number (1) Pressure u u (2) L* ax or Ax Ay Ax + p v C .
14 Letter R Re Re Rt Rx S Sc T t At u v v w Ax y Ay Az Greek Definition Universal gas constant Reynolds number Real Turbulence Reynolds number Reynolds number in x direction Source Schmidt number Temperature (1) Time (2) Parameter of a quadratic equation in Von Neumann analysis Time step size xdirection velocity ydirection velocity Velocity vector zdirection velocity xdirection mesh spacing Mole fraction in multicomponent gas ydirection mesh spacing zdirection mesh spacing a ala2 8 6ij C p v 4(1) 6 61 Thermal diffusivity or diffusion constant Parameters used in Von Neumann analysis Thermal volumetric expansion coefficient Kronecker delta Turbulent dissipation rate Density Kinematic viscosity Vpiscosity Phase change (2) Any general conserved quantity Velocity direction.1 v u Mesh configuration. 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 . tan.
15 Subscript cell cf DM f g i=l 2 3 4 L t m mix new ob old R t total US Cell Crossflow diffusion De Vahl Davis and Mallinson's Fluid Gas Air Hydrogen Steam Liquid Droplets Left hand side Liquid Molecular Multicomponent gas mixture New time step Obstacle Old time step Right hand side Turbulent Total calculation domain Upstream .
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. The four components. hydrogen. the fast blowdown stage and 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.( 1~111111_~. Some remedies have been proposed to deal with However. steam and liquid droplets are assumed to be in thermodynamic equilibrium and the relative humidity is assumed to be 100%. Some simplifying assumptions are made concerning the thermodynamic state in the slow mixing stage. although it may be less than 100% when there is no liquid component present. it is necessary to understand such problems. the fluid dynamic phenomena in the containment in order to justify those remedies.I~~U1i* 11*11044~1 1~. the response of the containment during an accident can be divided into two stages. From the hydrogen transport point of view. air. The error due to decoupling is negligible in a slow transient where the . The distinct feature of the second stage is a much longer time scale in comparison with the first blowdown stage.
Finally. The diffusion constant. and the total diffusion constant is the sum of the diffusion constants of those two mechanisms. of each component in multicomponent gas depends on the mole fraction of that component. Convection is assumed to occur as a homogeneous mixture. molecular and turbulent. The turbulent kinetic energy. . The turbulent diffusion constant is predicted by the kc model [45). 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. resulting in the same convection velocity for the four components. The turbulent kinematic viscosity. k. Diffusion occurs by two independent mechanisms. The molecular diffusion constant is predicted by Wilke's formula [73) for multicomponent diffusion and ChapmanEnskog formula [5) for binary diffusion. are determined by their own transport equations. Convection and diffusion are the central issues in physical modelling efforts of hydrogen transport.17 state change over one time step is small. and turbulent dissipation rate. 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. c. the phase change is taken into consideration to maintain 100% relative humidity.
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. The numerical diffusion has two different sources. numerical diffusion. the major obstacle in accurate numerical modelling of convection and diffusion has been the crossflow diffusion error which arises in donor cell treatment of the convection term in multidimensional problems.18 which is the diffusion constant for momentum transport can be calculated directly from k and E. the term. The effective diffusion constants of the two errors are of the same order of magnitude. The leading issue in numerical modelling of convection and diffusion is to minimize the error that occurs in the numerical solution procedures. however the latter turns out to be the dominant error source in most convection dominant problems. Therefore. The turbulent Prandtl and Schmidt numbers are assumed to be equal to one. Since the error usually appears as an additional diffusion. 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. Truncation error diffusion occurs in the flow direction while crossflow diffusion occurs primarily in the direction normal to the flow. There has been much debate on the numerical diffusion and many schemes have been suggested for the past two . has been used to describe the numerical error in general.
have been tested for recirculating flow problems. The stability conditions for each of the aforementioned schemes can be obtained by a Von Neumann analysis and turns out to be consistent with numerical experiments. However. The conservative property is essential in a long transient problem like the hydrogen transport in the containment. Raithby's [53) and S. The corrective scheme can be implemented with any of the explicit. ADI and implicit schemes. The mesh interface .19 decades. a recent review paper on this topic [65) shows that there is no scheme universally acceptable consistent with using reasonable mesh spacings and computation times. The corrective scheme is inherently better than the skew differencing scheme in that it is conservative and does not affect the simple solution procedure. although with different stability conditions. 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. Two implementation strategies for the corrective scheme. mesh point and mesh interface implementations. 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 schemes presently being used can be divided into two categories.
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. . simulation results compared with experimental data. [49) includes all the Development Laboratory) [6].
. installation of ignition devices. In addition to the pressure increase due to the primary coolant. The major safety concern is that of keeping the containment pressure below a certain level to prevent a large scale leakage of radioactive materials. After a while the slow mixing stage follows the initial blowdown stage and continues for an extended period of time. hydrogen generation gave a serious concern about the integrity of the containment in the TMI2 accident. a large amount of steam and water will come into the containment in the fast blowdown stage increasing the containment pressure. containment inerting. In order to justify the design of any mitigation system.g. 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. Thereafter. for dealing with this problem.. it is essential to understand the fluid dynamic phenomena in the containment.21 CHAPTER 2 PROBLEM STATEMENT AND SOLUTION SCHEME 2.1 Problem Statement Once a lossofcoolant accident (LOCA) occurs in a Light Water Reactor (LWR). the hydrogen has received much attention in the safety analysis of nuclear power plants. There have been some mitigation procedures suggested. e. use of flame suppressants and enhanced venting capability. .
phase change and heat transfer between the gas components and wall. a good computational tool has been required to predict the hydrogen concentration distribution.g. The most difficult aspects of this analysis are the complicated geometry and chaotic postLOCA conditions. etc._ __ _ __ _ I_ ~ 22 Since a numerical method is suitable for this purpose. thermal nonequilibrium.. some simplifying assumptions should be made to use the numerical procedure without impairing the acceptable solution accuracy. e. laminar and turbulent flows. Therefore. .
T) (2.2 Governing Equations The governing conservation equations are set up to describe the postLOCA fluid dynamics in the containment.6) Equation of state: p = f p (ii .Vp i + 4.3) Mas s diffusion: S+ V*vp i .S + + t Bui P au k aui + ax~ ) x k X axk gk Pr t t 'T xk (2.xk k 2 v t BT C2 + kg k Pr 3x 2k k t (2.4) Turbulence kinetic energy: Dk 1 t 3k 3 P 1 k ok x . I+ 3x 1 . Continuity: V'v= 0 pl + Vvv ] = Vp + Vv = 0 (2.23 2.1) Momentum: pgI + d + V0 (2.5) Turbulence dissipation rate: D 1 _ 1t StE xk Dt P Txk 0E ] + C t*( 1 p k ui xk ouk au. (2.t = VD. The physical implications of the conservation equations will be given with their basic assumptions and limitations.2) Energy: p[e e + Vvh] + = VkVT + e fg (2.7) .
t (2.15) Dt = v t (2.17) t = C k 2/C E (2.T) (2.T) (2.12) (2. The subscript i denotes the four components as . . = f (Pi.10) v = vm + Vt (2. Im + Dit it (2.24 Constitutive equations: 'N = f (p .9) D.14) m = fk (PiT) (2.11) k = .8) f (v) (2.13) D.k + k t m Dim = fD(Pi'T) (2. = D.18) p = Epi (2.16) kt = .19) Note that 1.
p. air. three momentum. four mass diffusion and two turbulence equations with ten primary unknowns. Therefore. The turbulence equations are written in terms of the Cartesian components in tensor notation.21) D4 = 0 (2. u.20) i = 2 = 0 43 + 04 = 0 (2. v.25 follows: i=l i=2 i=3 i=4 2. w. The mass diffusion equations are summed up for the four components. liquid droplet. T.i (4). there is one moreequation than is required. i=14. one energy. (2.k C for a three dimensional case.2.1 Continuity and Mass Diffusion Equations The continuity equation is redundant with the four mass diffusion equations and there is'an inconsistency between them.22) There are eleven conservation equations. 2. 3. . one continuity. H2' steam. The phase change occurs between steam and liquid droplets and the diffusion constant of liquid droplets is equal to zero.
25 and Eq.24) Now it can be seen that the continuity equation is valid if and only if the following Eq. at p= 0 (2.Vp. 2.23) The sum of the phase change terms is equal to zero.26 can be reduced to the following Eq. Eq. 2. 2.26) It is a reasonable assumption in a turbulent flow regime that the four components have the same diffusion constant.26 are satisfied.4 are consistent only if the temporal and spatial variations of the total density are very small. Therefore. . it is reasonable to assume that Eq. 2.) i + E 1 (2.26 .(Ep i ) a . ap at S+ Vp = E(VDiVp i ) 1 (2. 2. Vp = 0 (2.25) Vp i = 0 (2.27.1 and Eq. Since the concern is only for the slow mixing stage after a LOCA. 2.27) Consequently Eq. 2.) + V(Ep Jt i = (V'D.25 and Eq.27 remain valid throughout the transient. 2.
i + V. I at t + V [vp. LeaPi LeQ.i = V'kVT (2.30) Eq. 2. p + £e.ih. .2.iei) + V'[Ep. Be.27 2. Eq.2. is also of an approximate nature and the exact form is given in the following.ih i ] = V*kVT (2. (p. 2. (2.2 Energy Equation The energy conservation equation. E it p + e. = VkVT (2.29 to give the following. 2.31) The thermal equilibrium assumption is incorporated in Eq. 2.29) The second term on the left hand side of Eq.30 is substituted in Eq. [c pihi .3. It is clear that the major contribution to this term will come from the latent heat of the phase change.31 so that the internal energy and enthalpy can be written in terms of the temperature.28) The temporal term can be divided as follows. ae.29 denotes the energy change due to the concentration change in a given control volume.
32) where ei i = eg4g + e 4) Sefg g e. P (2 33) (2.T h.3 Basic Assumptions and Limitations The governing equations are based on some simplifying assumptions about the physical phenomena in the containment after a LOCA.32 become identical by defining average internal energy and enthalpy as follows.3 and Eq.T P P e Piei p ih p.33) h (2. = c p1 .c . Pic vi T c p p.28 (Picvi )~ +V"[vTZpic pi] + efg 9kVT = (2. .T 1 Eq.1 = c Vl. 2. 2. It is necessary to clarify these assumptions and their limitations for a safe use of the given physical models and solution scheme.34) 2.
The relative humidity in the containment is 100%. with no slip. 2.3. The four components are treated as being nearly incompressible and the phase change rate is moderately small. . The turbulent Prandtl and Schmidt numbers are equal to one. 2. The four components all have the same convection The liquid droplets move with the gas components velocity. 5. The temporal and spatial variations of the total density should be negligibly small because of the incompressibility assumption. The relative humidity may be less than 100% when there is no liquid component left. Dt and k t over one time step should be small because the turbulence equations are decoupled from the velocity field calculations.29 2. The change of the diffusion constants. 4. 3. 3. The four components are in a state of thermodynamic equilibrium.1 Assumptions 1. vt. 2.3.2 Limitations 1. The phase change rate should be moderately small because the solution scheme decouples the phase change term from the energy conservation equation.
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. continuity/momentum equations set and other scalar transport equations set. 2.1Eq. 2. p. When there is no liquid component left. 2. The phase change is assumed to be zero in this step. Eq. T.7.5. explained in detail in section 2. k. It solves the . . is used to update the reference pressure and reference density which is important in calculating the buoyancy force. In the third step phase change is taken into consideration to maintain 100% relative humidity.i Since it is too difficult to solve the whole system of the equations and obtain a consistent solution for all the unknowns.4 Solution Scheme Eq. The SMAC scheme is In the second step. . the obtained velocity field is substituted in the scalar transport equations. In the first step the SMAC scheme is used to get the convergent velocity field with zero divergence for every mesh. u. the conservation equations are decoupled into two sets. 2. Pi(4). energy equation and two turbulence equations. The equation of state.7 are a coupled set of equations with primary unknowns.30 2. which are the four mass diffusion equations. v. E. the relative humidity may be less than 100%.
36 and Eq.e. However.ax 2. The continuity and momentum equations are given in two dimensional Cartesian coordinates in the following.37) Py The vorticity conservation equation is derived from Eq. 2. (Eq. ay (Eq. it is possible to modify the SMAC scheme to accommodate a slight compressibility effect. Continuity: au ax av ay 0 (2.36) ydirection momentum: av av ua av 1 + pg + a y x2 Nv 2 (2. Since the continuity equation used in the SMAC is an incompressible form.37)y (2. the divergence of the velocity field vanishes everywhere. which can arise with net inflow or outflow boundary conditions. 2.36) .37): .1_ _~ 31 continuity and momentum equations to get the velocity and pressure field in an iterative way.37 as follows.. i.35) xdirection momentum: au au + u+v at ax au ay 1 p ax + Pgx + a 2u 2 ax 2 2au + 2 2 (2. 2. the SMAC scheme can be used only for incompressible flow calculations.
41) where fn and gn are the explicit quantities at time step n.37 as follows. the socalled tilde phase and pressure iteration.32 a C 2 a2 + u+ V = v+ ) at .36 and Eq. ay 2.x ay ax 2 ay2 where au av (2.38) =ay ax The Poisson equation for the pressure field is also derived from Eq. 2. In the tilde phase the approximate velocity field at time step (n+l) is obtained from Eq.37 as follows. ax 2 2. (Eq.36) + ~(Eq. 2.37): a2 = 2 ax 22 axay a 2 ay 2 (a2D + a2D aD +1 2 at Re ax 2 ay (2. u n+lu n u At n+l At n n 1 Apn+l A + n 1 p Ax n+l + gn 1 Ap p by (2.+ y ax ay The basic idea of the SMAC scheme is to separate each calculational cycle into two parts.36 and Eq.40) S (2. 2.39) where av au D= . 2. .
V2p = f(u. follows.41 are still unknown. 2.40 and Eq. 2. 2. 2. 2.43) ax ay It can be seen from Eq.42) av .O 0 (2. The equation of vorticity transport. a guessed pressure field or the pressure field at time step n is used to start the iteration.44) Eq.38. shows that the vorticity is independent of the pressure field.45) .v) D (2. 2. Eq. The impor tant point is that un + l and vn+l calculated from Eq. the following relations hold after the tilde phase.44 is given in the following.39 may be rewritten as A finite difference form of Eq. 2. Therefore.40 and Eq.39 that the pressure field is not right due to the fact that the divergence of the velocity field is not equal to zero. au ay au av ax  (2.33 Since the pressure field in Eq.41 have the right vorticity. _ Pi+j2P +P iPij+ pij f(uv)  Dn + l D n D D Ax 2y t (2. 2.
j) 26p( 1 Abx 6p = h + _) Ay 1 At( A + 1 Ay 1 Dt D (2. no more correction will be made. equal to zero. The velocity correction formula is derived from Eq. Once the zero divergence velocity field is obtained. while the continuity equation has the following compressible form. S+ at V"(p) = 0 (2.47) ) The pressure correction is given in terms of the residual divergence for every mesh.49) . (6u) L =  pAx (6u) R = t6p pAx (2.36 and Eq. The slight compressibility means that the NavierStokes equations may be solved safely with the incompressibility assumption.48) The SMAC scheme can be extended to a slightly compressible fluid flow.34 The pressure correction formula is obtained by conn+l only and putting D.. sidering the cell (i. 2. 2.37 so that it does not change the vorticity implemented in the tilde phase.46) (2.
51) v p where m.51 and Eq.50) The two terms on the right hand side of Eq. D = V'v 1 p p at V p (2. 2. 2. Now the problem is how to converge to a predetermined nonzero divergence field. This can be done with the following modification of the pressure correction formula of the SMAC scheme. The associated physical assumption in Eq.uniform over the whole containment.50 may be given from the previous time information and boundary conditions.52) (2.49 the nonzero divergence may be derived as follows. 1 .52 is that the pressure perturbation due to the inflow at the boundary propagates throughout the containment instantaneously and the pressure and density increase is. 2.p_ p at 1 min/V At n p n n (2. 2. V: total volume of the containment.35 From Eq. : in n =Vp v  total net inflow of mass at the boundary during the time At. .
53) where D O is given by Eq. 2. The velocity correction formula remains the same as Eq.(DD 1 1 + ) 2At( Ax Ay (2.50. . 2.48.6p = .
air. As collisions occur more frequently.1. proper modelling of the latter is important because the molecular diffusion of hydrogen is significantly greater than those of other gases and also because there may be a laminar flow region in the containment. 3. molecular and turbulent. The thermal conductivity and viscosity are also diffusion constants in a broader sense for momentum and energy transport. Convection occurs as a homogeneous mixture resulting in the same convection velocity for the four components. in the containment after a LOCA. components.1 Molecular Diffusion Molecular diffusion occurs by the collisions of gas molecules. the process Since there may be three gas of diffusion is also increased. hydrogen. the diffusion constant of each component is calculated separately by Wi~k's formula [73] in Eq. steam and liquid components are transported by convection and diffusion in the containment. 3. which should be modelled independently. air and steam. It gives the diffusion constant of each component in terms of . Diffusion occurs by two different mechanisms.37 CHAPTER 3 DIFFUSION MODELLING The hydrogen. Although turbulent diffusion is greater than molecular diffusion by a few orders of magnitude.
0018583 1 + M A MA MB (3. 3. Eq.2) poAB D.1) where D.2. The binary mixture diffusion constants are calculated by the ChapmanEnskog formula. stants are given as follows.. The property of air will be replaced The resulting diffusion con with that of nitrogen gas.: Binary mixture diffusion constant between the ith and jth components. 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.AB where DAB: AB Binary mixture diffusion constant in [cm /sec] . T3 D = 0.: yi: Diffusion constant of the ith component Mole fraction of the ith component D.
In order to calculate the para meters aAB and 0DAB in Eq.lll~*. For nonpolar gases OAB and ODAB are calcu lated by Eq. ap = np £ (o n + o p) EE e (3. =n [ p n = [+ S n . 3. the following relations are required.5) 2 n .2. 3.3. (3.3) CAB = AB For the pair of polar and nonpolar gases a correction factor is introduced to account for the polarity.AB: Dimensionless function of the temperature and intermolecular potential field for one molecule of A and one of B. OAB = (o A + oB (3.. P n ]. The error range of the ChapmanEnskog formula is reported to be about 610% [5).4) np = np 2 where the correction factor 4 is given by.UIIII ~ II~PI^ICIL1I n~s^  ^ 39 T: p: aAB: MAMB : Temperature in [OK] [atm) Total pressure in LennardJones parameter in [A] Molecular weight QD.
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. 3.= 1.40 The parameters in Eq.6 x 10 The fitting functions for the binary mixture diffusion constants are obtained by the leastsquarefit method.6810 x 10 . H2 a n = 7.5 are defined as follows.2 E/K = 380 0 K a = 2. t* .9035 = 4.6947 as .6 T1.65 A The polarizability is given for the hydrogen and nitrogen (air) components.9492 x 10 D ah D 1 T . 5 = 4.9 x 10 25 25 25 [cm3 3 [cm ] : N2 : an = 17. The containment pressure is assumed to be1 atm.
67 x 10 where : T: a: Svi : o2 2 (3. i (3. = n i=l y.lii.6) The viscosity and thermal conductivity can be considered as diffusion constants for momentum and energy transport and calculated similarly as follows [5).10) Viscosity in [poise) Temperature in [K]) LennardJones parameter in [A) Dimensionless nTumber .__Li~n_~PIIII ~.8144 (3. n mix = Yipi n i E j=l Yj ij (3.7) k.lllI_ 41 Dhs = 2. = 2.k.8) j=l where ij 8 Mj j ij M The viscosity of nonpolar gases may be obtained by the following [5).8916 x 10 where T: D: [OK1 [cm 2 /sec] 5 T 1.
E.42 The viscosity of steam is. c (3.009 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. 200 C There is another formula suggested for the viscosity of multicomponent gas in the following [5].11) for the Prandtl number of Pr = c (3.[5].2 Turbulent Diffusion The diffusion process is greatly enhanced by the turbulence of fluid flow. Wsteam = 0. 2 n Yi umix = C 2 i=l Yi + 1. seems to be the kc model originally developed by Launder and Spalding [45).y RT k=l PMiDik k i There is a useful relationship nonpolar polyatomic gases. The . 3.385 nn y. The effective diffusion constant is therefore the sum of the molecular and turbulent diffusion constants. and turbulent dissipation rate. k.12) where c P is the specific heat per mole at constant pressure.
13) (3.u. = u.' For example. The effect of turbulence appears as the additional terms due to the product of fluctuations which do not necessarily cancel out.43 turbulent viscosity is given directly in terms of k and C.' etc. 34. (3. . 29. These terms are treated by the eddy viscosity concept of Boussinesq and eddy diffusivity concept which are given in the following.= v ( Dt + ui j u P= Dt ax x 2 k6 ij 3x_) 3 i (3.' j T'= a t x.14) .u. etc. + Pi Pi = Pi + p. momentum and energy equations by decomposing the variables into mean and fluctuating parts and averaging the resulting equations. + u.15) .1 Derivation of Turbulence Equations [15. 45] The transport equations for the turbulent kinetic energy and turbulent dissipation rate are derived from the continuity. u. 3. u.2.
t + at r a (ruk) + (wk) = [ p r (r oka r ) + (t ) az ok a au aw + ) r aw au 2t + 2( + 1[2() 3 3r p + 2 + u2 2 ] .2. 3.16.16) . 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.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 (3.44 The transport equation for u. A two dimensional cylindrical coordinate form of the kE model is introduced in Eq. j=i.2 ke Model The transport equations for the turbulent kinetic energy and turbulent dissipation rate form the basis of the kE model. 3.' u is transformed to the turbulent kinetic energy equation by the contraction.
17) The suggested values of the constants are given in the following table.17.3 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. 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. = C pk 2/ (3. 3. it is difficult to determine whether a region under consideration is in a turbulent or laminar flow.metry. 3.92 ok 1.16 as follows.3 The molecular effects can be included in Eq.0 o 1. C 0. 3.45 The turbulent viscosity pt is given in terms of k and E in Eq.09 C1 1. Since the kE model is applicable to fully .44 C2 1.
0 a ui ( 2 p ax ax 2 ) Dk Dt 1 p ak [(.16. CP = CP0 exp[2. D Dt p ax k [(t Dt+ ) e k + C k i + k x ) au axk C 2 E2 2 t 2.0 .19) = pk2/VE .0.5/(l+Rt/50)] C2 =2 where R 2 0 [1. it is required to extend the model to laminar flow regime or to switch to laminar models according to some criterion. 41). The model presently being used is just to use the sum of the molecular and turbulent contributions for every region in the containment.18) In the above equations C 1 1 ok and a retain the values assigned in the ordinary kc model.2vak 2() k 2 (3. 3.46 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. while C1 and C2 are to vary with turbulence Reynolds number. Eq. The following kc model is a modified form by Jones and Launder [40.lit k + l ak aui i + t (ax k + xk auk axi axk e . Rt.3 exp(R2t ) (3.
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)
6 from Eq. It can be extended to a two dimensional form as follows.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.x+a) and [y.y+b) = f(x. we obtain the error at time step (n+l). 4. Eq.7. 4. n . n+l i+f EXACT n+l i FD n i+f  4P n a n+g i + At [ u( x ) i+a (4.3.8) 3 @n + u()n + a( ax i+b ax 4 n+g _ n )n+g . af f(x+a. .y) + (f) af (4.9) 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. 1 < c < n 2 4 + i Ax 2 11 4 2 1 i+d < d < 1 The mean value theorem.y+b]. 4.i+c T i+d ax where the following relations have already been used. n+g 4 i+ Ax n i Ax 1 xji+ n i+l .51 Subtracing Eq. is for a one dimensional case.a(?) ) 2(.
D +aAx(cd)Atax 2 +agt . Although the error terms in Eq. 4. The terms A. E . C and E are truncation error diffusion terms while the terms B.A fAxAt axat 2 B utx(a+b) . they may represent the errors over many time steps in a stable scheme. . 2 n+l (4i ) EXACT  (4+) FD = i n+1 f )2 2fx 1f 2 ) (f 2 2 .10) The derivatives with respect to time and space in Eq.8 to the following form. 4.10 are at some appropriate points in the domain [xi .9 reduces Eq. 4.10 are based on the assumption of perfect information at time step n.xi+ ] and [tn tn+).C ax 22 +At g t .52 Now Eq. 4. In a stable scheme the error introduced at a certain step decreases in its absolute magnitude in the following steps and most of the error occurs between neighboring time steps.tax .F (4. D and F are time derivative terms which become negligible for slow transients.
12) and C 1 =N cL 1 e C2 ecL 1 .1 Truncation Error Diffusion in a One Dimensional Problem The truncation error diffusion in a one dimensional problem is analyzed for a steadystate case with no source. 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. cL (4.11 is given as. u ax a22 (4.11) ax Consider a case in which the domain is divided into N equal meshes and the mesh size is Ax. 4. The analytical solution of Eq. 4= Cecx + C2 where c = u/a .53 4.1. 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.
13) where the velocity u is positive.14 may be recast to Eq. .14 is the cell Reynolds number or cell Peclet number according to whether the diffusion constant a is the kinematic viscosity or thermal diffusivity. Equation 4.15) Equation 4. i+  i = (P+l)(4)  4)i (4.14) The parameter P in Eq. where P =u a  l i24 + ii (4.i + 4i2 Ax = a (4. i u x i+l2.54 Equation 4.15 and solved for in subsequent procedures.16 is easily obtained from Eq. i+l i = (P+l)( 1 . P(4).15. 4.11 is finitely differenced using donor cell scheme for the convection term in the following. 4. (4. 4.40) . Equation 4.13 can also be solved with the given boundary conditions by reducing it to the following form.16) .
4. we obtain the solution for 4. 2 1 = (1+P) (4 0) 4' N 92 = N2 (1+P) 2 (  0) = N.41 = N ) (1+P) [(1+P) N1  2) (4. i  0 = 4. 4.17) 01 can be expressed in terms of 40 and 4N N N by Eq. 4.12 and numerical solution Eq.(1+P) P (1+P) 0 + N .~l*IIIIX1~~^ ~ ^I ^~Xi 55 Then. as.19 into Eq.'N1 (1+P) N1  ) 0 N .13.20 have been obtained without any approximation for Eq.I_. (1 0 ) is ) P( N ¢0 (1+P) given as follows.18) (l+P) + 1 Therefore.17.16. 4.LI II .20) Now the analytical solution Eq. 4.1 Inserting Eq. 4.1 (1+P) N 1 (4. (1+P) 1 . Both the analytical and numerical . 1 (1+P) i .~LI LIII~. 4.19) (1 .11 and Eq.
13.24) It is also possible to go through the similar procedures with a central differencing form of the convection term in Eq.+1 i + ) il 0 (4. 24 j. 4. 4.20 is also substituted for 4. Eq. i+l 2Ax il S S2ax 4) i+l 24i + i 2 i (4. Pe = in (1+P) (l+P) iIp2 N (1+P) .12 is substituted for 4i Eq.22) Comparing Eq. + 24 S i i+1 i1 =e e l NP  1 (e P P 1) 2) ) ( (e 0 (421) The finite difference solution Eq. 4.23) (4.25) (4.13. 2 ii P ( (1+P) (1+) N _ 1 N 4). 4.56 solutions will be reinserted to the finite difference equation. 4. in Eq. e (il)Pe (ePe1)2 NPe .25. 4. The in analytical solution Eq.25) .21 and Eq. 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.1 (4.13. and the effective diffusion constant for the numerical solution will be obtained.1 e Therefore. 4.
57 The effective Peclet number for the finite difference solution of Eq. 4. Pe = kn (2P) 2+P (4.26 may be a good indication of the truncation error diffusion occurring in a general one dimensional problem. 4. Although this approach is approximate and overpredicts the truncation error diffusion. As in the previous section the velocity u is assumed to be positive and constant. The convection terms in .2 Another Approach For a One Dimensional Truncation Error Diffusion The truncation error diffusion in a one dimensional problem can be evaluated in a different approach. Therefore. Eq. 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 4.27) In general. uLx D De Pe  (4. 4.24 and Eq.1.26) where 2 < P < 2 The effective diffusion constant De can be readily obtained from the following relation.25 can be derived as follows. it helps understanding the origin of the truncation error diffusion. 4.
29 from Eq. u i Ax i1 u i+ a4 u() ax u(a) ax (4. 4. 4.29) i+ The truncation error diffusion can be quantified by subtracting Eq.24 and Eq. 4. the diffusion is dominant over the convection.. 4. 4. Therefore. 4. 4. It can be shown that as P goes to zero.29 hold only when the Peclet number is large. i. If the Peclet number is small.e. the truncation error diffusion constant for a convection dominant problem may be given approximately as follows.28) Ax (4.28 and Eq.5 may be approximated as follows.e.30) Both Eq. 4. ae u( x where 0 <f < ae + u() ax i+  u~x a2 ) . the profile of 4 will be linear and the truncation error will be reduced to zero.58 Eq.26. Pe also goes to zero resulting in no diffusion error in Eq. D uAX 2 (4.4 and Eq.( i+f 2 (4.31) ND .28. i. the convection is dominant over the diffusion.
two dimensional.1.33) . A steadystate.32 is finitely differenced using donor cell scheme for the convection term. It will be shown that the truncation error diffusion in the direction normal to the flow cancels out and there is only a flow direction component left. Since the gradient of 4 in the flow direction is negligible in a convection dominant problem. u i 6x ilj + v i by ij_a(i+1 2 4x i + il1 +ij+l 2i 2 + 0iji by (4. one is the truncation error diffusion and the other is the crossflow diffusion.32) = x 3y Equation 4.59 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. conservation equation with no source term is given in the following.3 Truncation Error Diffusion in a Two Dimensional Problem There are two sources of numerical diffusion in a two dimensional problem. the total diffusion quantity of 4 due to the truncation error is not significant in comparison with that due to the crossflow diffusion. 2 u + v 2 + +( ) (4.
33.j+ 2 2 ay 2 a2 2 A Ax U(a 2 ) + bay U( 924 )+ Ax 82 2 axay 2 ax2_I = AX Ax A 2 axay va a yyaxa 2 ay 2 S[uAxax + (uAy+vbX) +v = .x[cos + (cose tane 1 + sine) + sine tane ] ax xY (4. transformed to the coordinate ( .v( ay ) ij .34) where tane = Ay/Ax tane1 = v/u The coordinate (x.35) . then the truncation error of the convection term is given as follows.n) ( = x cos6 + y sine I = x sine + y cose (4.60 The approach in section 4.2 is applied to Eq.y) is by rotation. 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 .j+ + v(a) ly i+h. 4.1.
y v e Sx Then Eq.61 where 5 is the coordinate in the flow direction.37 shows that the diffusion component in the n direction (normal to the flow direction) is zero and the cross differential term also vanishes when e is equal to e1 Therefore. e is equal to 0 or 7/2.34 will be transformed to the coordinate system (t. 2 2 2 + a = ( 2 ) (4. the problem is basically one dimensional and the truncation . 2 (A)  [(sine tane1 + cose) 2 + a2 (sine + cosetan6 1 ) a + (0) 4 an U" (4.36) where U = /u2v 2 Now Eq. 4.32 is reduced to the following form.37) Equation 4. 4. 4.n) given by Eq.35. dominant truncation error diffusion occurs When only in the flow direction.
4. In this section the origin of the crossflow diffusion will be illustrated and the corresponding diffusion constant will be quantified so that they can be used in the corrective scheme of section 4. 4. 4. It is entirely different Recently this addi from the truncation error diffusion.62 error diffusion occurs in the flow direction as in a one dimensional problem.1. The crossflow diffusion comes from the multidimensionality of the problem. which shows a single mesh with pure convection. [67. 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.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.l1(A). In Fig. The hot fluid will come out of the top surface and the cold fluid out of the right surface.3. therefore it exists only in two or three dimensional problems when the flow direction is not aligned with the mesh configuration. In the donor cell scheme Fig. tional false diffusion was explained by Patankar [52] and Stubley et al.68) and it was clarified that this is the dominant source of the error in most multidimensional problems. 4. 4.1(A) is transformed into .
63 Hot Mledium Hot 7 Hot Cold (A) Cold (B) Cold (C) Fig.1. 4. Illustration of the crossflow diffusion in a single mesh with pure convection .
in The numerical results Fig. Then homogeneous mixing occurs in the mesh. diffusion occurs in multidimensional donor cell differencing of the convection term. of the value of 4 causes the crossflow diffusion. and the sum of the values This distribution along the points A1 . A 2 . . 4. 4. at those points is exactly equal to 4.1(B) where the velocity components normal to the surface are considered. 4. an appreciable amount of false program user. 2 (lp)[l+p+p + . ] = (lp)4 = where u . The value of 0 at the point A should reappear at the point X because there is no physical diffusion.64 Fig.and intermediate temperature fluid will come out of the top and right surfaces. Another illustration of the crossflow diffusion is given in Fig. CAB which looks like a long onedimensional rod. 4. . This is a pure convection problem Consider the fluid element The fluid with no physical diffusion. . .2 which shows a row of meshes aligned in the xdirection. In the donor cell scheme it is distributed all .&x u v ax Ay The effective diffusion constant for the crossflow diffusion will be calculated in the two dimensional Cartesian coordinate in Fig.3.1(C) by a Therefore. A 3 . 4.1(B) are again interpreted as Fig.
Illustration of the crossflow diffusion in a row of meshes aligned in the xdirection with pure convection .65 (1p)4' x p (1p) 2 p (lp) p3 (lp)4 p4 (1p) A5 Ay A Ax 2 4 y p t* U where p = Ax Ay Fig.2. 4.
Geometry for explanation of the crossflow diffusion in a two dimensional Cartesian coordinate .66 4. V A2 e A. AY B Ax t^ t where t t0 = tane = v/u u/Ax v P U = 3/U tane= Ly/Ax Ax by Fig. 4.3.
The diffusion occurs both to the The diffusion to the right hand side is considered first in the following. Diffusion A t1 . right and left hand sides. k2 = Ax sine/sin(el+8) k3 = Ax sine 1 /sin(e 1 +8) After substituting appropriate expression for each term. 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'.67 element was on the line CAB at time t in the velocity direction. which shows that the value 4A diffuses out along the fluid element. the following right hand side diffusion constant results.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 . 4. In the donor cell scheme it is divided into two portions. v.4. sine cose cos= sx Uacos61 sin(O+e ) I DRHSS . Gradient) A /k 1 =A of 4 Current 2 2U of 4 Therefore.
4. 4. Ya.ct t2 t3 Fig. Geometry for explanation of the crossflow diffusion in a rotated two dimensional Cartesian coordinate ..68 ti .
38 is the effective diffusion constant when the diffusion is confined to the one dimensional fluid element. 82 x 9x 2 2 By 2 2 The current on the one dimensional fluid element can be factored into the x and y components. sine cos6 D = UAx sin cos8 x sin(e+e 1 ) 1osel 2 sine cose sin 81 D = UAx y sin (8+8 ) cos1 1 (4. sine cose RHS LHS cos1 sin(+8) (4.39) 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.39 is identical to the following simple expressions.40) . The crossflow diffusion constants are the ratios of the current and gradient of 4 in the x and y directions. There are two diffusion components in a two dimensional problem as follows.38) Equation 4. D D = uAx(lp) = VAy p (4.
38 should be differentiated with respect to 6 in order to find out the velocity direction with maximum crossflow diffusion.41 shows that DRH S (or DLH S ) is maximum at an Tr7LHS angle of 61 < 6 e between 4 and e1.42 underestimates the crossflow diffusion as shown in Fig.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. that is to say. Now the analysis will be extended to a three dimensional case in Fig. 4.42) Equation 4. Eq. 4. 4.42 give the same result.40 and Eq. When both 6 and 61 are equal to . De Vahl Davis and Mallinson [24] also derived the effective crossflow diffusion constant and their result is given in Eq. In the donor cell p' y scheme the value 4 is divided into three portions px 0 .70 where p = u Ax u v By Equation 4. 3 0 when tan 6 = tan6 1 .5. 4. 4. D DDM = UAlxy sin2e= 3 3 4(by sin36 + Ax cos36) UAx tan6 1 sin26 3 4(tan8 sin6 3 + cos3) (4.6. < 6 < 81 or < . In Fig. (4.42. RHS D = It can be shown that. 4.41) Equation 4.
Comparison of the prediction formulas for crossflow diffusion constant by De Vahl Davis and Mallinson and Huh . 0.0t Dcf UAx de Vahl Davis & Mallinson Huh 81=8C 1.0 150 300 450 600 750 900 Fig. 4.5 I I 0.71 Cross flow Diffusion 2.5.5 80 \ % % 4 1.0 0 0 .
p Fig.72 z C V Az X 0 o a X AY Ay B y x z C p .6. Geometry for explanation of the crossflow diffusion in a three dimensional Cartesian coordinate . 4.
BX. menon.43. 4. X(£ 4cosa. n C denote The coordinate of the Any point on the plane ABC should satisfy the following Eq. S+ Ly Ax + Az = 1 plane ABC (4. XB.45) The lengths of AX. cosa cosB cosy (4. point X is given as follows. the following relation holds.44) 4 Therefore. and C instead of at the point X.73 z occurring at the points A. CX which are denoted as £i' z2' k3 can be obtained as follows. S= 4 + y z zAx 1 cos+ + cos Ay Ax + cosy Az (4. B. t4 cos8.43) Since the point X is on the plane ABC. XC. along the directions XA. This is the socalled crossflow diffusion phenoThe currents due to the crossflow diffusion are . £4 cosy) nB .and nA unit vectors along those directions. £1 = AX =/(x4cosa)2 + (4cos)2 4cos) 2 = Ax 2 + 42 2Ax£ 4 cosa .
OX (4. z directions. n C can be obtained in terms of their components in x.47) etc.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. Therefore. nB. (Ay4 cosS). y. I (Current)xAi XAI = px Px C1 1 Lt .48) The currents in the directions nA . 4 cosy] 1 nB = [i4cosa.  n A = XA +1A  OA .46) + The unit vectors nA . 2 o C 14cosy] = 1  £ 4 cosa. Z 4 cos8. nB' n C can simply be given as the products of the transported quantities and velocities.
50) Now the x direction crossflow diffusion constant may be given as the ratio of the current to the gradient of 4 in the x direction. (4.49) The gradients of 4 in the x. z directions are given as follows. (Grad)x = 4./ Ay 4/6x Q/by (Grad)z = 4/Azl (4.51) p Ax D x = lt (Axk 4 cosa) + p Ax . y. Px D (Ax£ + Py D + nC(4 (Current)x 4 cosa) c osa) At ( 4 (cosa) (Grad)x = 4/Ax Therefore.75 I(Current) XB = Py p2 t I(Current) XCI where At = k /U z '3 At (4.(4 pz x co s a ) ct 4 + at ( 4 cosa) 4 ./ Ax (Grad) = 4.
most important ones may be the skew differencing.54) 4. etc.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. tensor viscosity method. Dy = vLy(lpy) Dz = wAz(lpz ) The rotation of (4. finite element method.1 = uAx(lpx) (4. indices will also give the same result. . This section is a review of the schemes that The have been used to eliminate the numerical diffusion.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. p Ax D = ux[ £4cos x .53) The crossflow diffusion constants in the y and z directions can be obtained in the same way.52) Inserting the expression for At given by At = £4 /U. although none of them has been successful enough to be accepted widely. we can simplify the result as follows.
Basically the inclusion of four additional corner points contributes tomore accurate numerical modelling of the convection. the interpolation should be done between neighboring points to obtain the value at the upstream point.~_ 1 i_.. The donor cell treatment of the convection term results in a 5point relationship in a two dimensional problem. Since the upstream point in the velocity direc tion does not necessarily fall on the mesh point where the quantity under consideration is defined. The corrective scheme is to reduce the diffusion constants in order to compensate for the additional false . but just appreciably lower than that of the donor cell scheme.3. S. but directly along the velocity direction.^ILli. skew differencing and corrective schemes.4 fall in the category of a corrective scheme. the interpolation between neighboring points may still give some crossflow diffusion.. However. The crossflow diffusion is not eliminated entirely in the skew differencing scheme.1 can be categorized into two types. 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... In the skew differencing scheme the 5point relationship becomes a 9point relationship.
thod LLxW'en doff  21 x11 80 x 400 Table 4.tnce Vuer:o: upwind Yes Yes Yes Yes 2) x II 21 x ]1 21 x 11 21 x 11 41 x 21 41 x 21 21 x 11 81 x 41 28 x 14 Se Tzb)t 2 Up'nd ft(e elemmnt !./LLUE Huint/QUICK Wilkes/LUE Lillimpon/SUD LiiL. £Ebzhau t aL/PATANKAR Lilinpion Orlandi Priddin Wada cit l.1QUICK t ETovh ci .j.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 .ronLUD Lillinion/VUDCC Nuana..ordc. Grandotto Hickmott Firs. List of the schemes for numerical representation of advection presented at the Third Meeting of the International Association for Hydraulic Research.1. Esposito Glas and Rodi Huffenus Lnd Khali. 1981 [65) . upwind f~iite di Teretn atl.thod of chaaceriics Fiite anuryic method Tenso.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 . viscosity method Sel. Wilkes Elbahu ct l.hod Mesh *txLnsformation rr.aptjve me.inte Conribo:r/me.e elements Clifft Dontcl t al.ad.Highe cdtr upwi:nd :"'ie Cdi't.78 Chss' iction Cen:ered .
and D z in Eq. 4. The corrective scheme is valid only if the effective numerical diffusion constant can be predicted accurately. Since the crossflow diffusion is dominant over the truncation error diffusion and the effective diffusion constants for the crossflow diffusion can be predicted theoretically. 4.3.79 diffusion.1 Raithby's Scheme Raithby's scheme [53] is an Eulerian type of approach. 4 = C1 + C2 n = C1 + C2 (  (4. the corrective scheme can give a numerical solution which is almost free from numerical diffusion. although the validity of its correction formula is not clear. The tensor viscosity method is one example of the corrective scheme. 4. The correction formula The diffusion con used here is Eq. It is therefore a conservative scheme. 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.54. where the balance equation is set up for a given control volume by considering convection and diffusion at the control surface.54. 4. stants in x. 4.53 and Eq. y and z directions will be reduced by the amounts of D .55) where n is the distance normal to the flow and U = /u 2 +v 2 .. D.53 and Eq.
80 Fig. A segment of. control volume and notations in Raithby's scheme . 4. the calculation domain showing grid lines.7.
4.= U ax dy SC (4. The convection term is differ enced in the velocity direction between mesh points. Although this scheme is nonconservative.2 Skew Differencing By Huh Skew differencing scheme here treats the convection term in a Lagrangian way. 'ij US (4. The inclusion of the four additional points also complicates the structure of the coefficient matrix in an implicit scheme.56) where U = /u is the coordinate in the velocity direction and 2 +v 2 .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.57) where the upstream value €US may be calculated by . which may be too complicated and time consuming for a practical purpose. 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. u 4 + 'v. it is much simpler than Raithby's scheme and becomes conservative in a unidirectional flow.3.
53 and Eq. Some of the results of this skew differencing scheme is given in section 5. The validity of this scheme is based on the fact that the crossflow diffusion can be predicted accurately.3.4 Corrective Scheme By Huh The corrective scheme is to use the reduced diffusion constants. Dx = uAx(lpx) D Yy = Ay (lpy ) The expressions for . DD y . the tensor viscosity [25) T = Lt uu u2 The expression for is given in the following. 4. D . DD z in x.58) t2U= Eq.2.54. uv (4.82 interpolating two neighboring points. the crossflow diffusion constants D in section 4. 4.3. z directions to compenx sate for the crossflow diffusion effect.3 Tensor Viscosity Method The tensor viscosity method treats the numerical dif fusion constant as a tensor quantity. 4.2. 4.58 for the numerical diffusion constant is apparently not consistent with the results in Eq. D are derived .3. DDx. y. 4.
1 Mesh Point Implementation The mesh point implementation is to assign the crossflow diffusion constants Dx .+ . which is used with the cell dimensions to get the crossflow diffusion constants.+ w Ay Lz x etc. the mesh point implementation and mesh interface implementation.. In a coarse mesh or recirculating flow field the latter implementation strategy is definitely preferred. Dy .4.59) v u . 4. After the crossflow diffusion constants are assigned to every mesh point. 83 D z = wAz(lp z ) where px = u (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.Dz at the center of a mesh where all quantities are defined except the flow field. they should be averaged again between neighboring mesh points to calculate the crossflow diffusion current at mesh interfaces._IIU__lpULI~~I___~lLlt i~Lli ~i ~iL~pn~sl. .3. Two implementation strategies have been tested so far. The velocities at two boundary faces are averaged to get the representative velocity.
There are three possible cases that should be treated independently in a two dimensional case. The right interface will have the crossflow diffusion constant (Dx1+Dx2 ) and the top and bottom interfaces. U1 and U2.4. so that they are proportional to V1 and V2.59. For the first two cases the cross flow diffusion constants are zero because the velocity vector is aligned with the mesh configuration. 4.2 Mesh Interface Implementation The mesh interface implementation is to assign the crossflow diffusion constants directly to the mesh interfaces where the crossflow diffusion current should be calculated. The two velocity sets (Ul.3. . 4.84 4.8. The mesh interface implementation is always recommended although it has a more complex program logic than the mesh point implementation. as shown in Fig. The velocity U is split into two components.Vl) and (U2. and the resulting D x and D y will be assigned to each interface. U and V. Dyl and Dy 2 .V2) are used to give (DxlDyl) and(Dx 2 'Dy2 ). The third case is expected to have a flow split as shown in the figure. Note that the crossflow diffusion constant for the interface with an inward velocity does not have to be considered in that mesh because it will be considered in the neighboring mesh. can be directly used with cell dimensions in Eq. The same logic can easily Appendix C includes be extended to a three dimensional case. For the second case the two outward velocities.
8._~I~~ _~~~~_I__~__~~ __ 85 (1) (2) (3) Ul U2 V2 VI U U V14V2 VI+V2 V2 U2 = UV2 U = Ul + U2 TV2 Fig. Three possible flow configurations for mesh interface implementation of the corrective scheme in a two dimensional problem . 4.
2. 1. 4. Eq. 3. for the general conservation equation.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. Implicit and ADI (Alternate Direction Implicit). The pros and cons of the three schemes will be compared in the following four viewpoints.) = i1 n D Ax ( i+1 2i l n n + n ) + iS i1 (4.4 Comparison of Explicit.86 the fortran program for calculating crossflow diffusion constants in an arbitrary flow field.62) .60. 4. which is a parabolic partial differential equation. conservation physical constraint accuracy stability 2 ax 2 2 y 2 +ux +at + 7 + T7) + S sz (4. Explicit. 4.60) Explicit (lD): n+l I n i At + u (i i x n i. Implicit and ADP schemes There are three typical solution schemes.
2 Ax n+l 2 * ij v n+l ) ij n+l Ay 13j1 n+1 ij+l n+l 1j1 At Ay2 + S (4.63) 4.87 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. or affected only by the source term contribution (SO). when the net inflow or outflow at .4.. + 4).1 Conservation The conservation principle is that the total amount of 4 should be conserved (S=0). n+ n+ + 4n+ n+l + D ij+l  2 n+1 + 4n+l i i1 + Ax2 ADI (2D) : * Ay 2 (B) n *i * 1 j + 1 1 i+l  At _ S +Ax 24.
there are some nonconservative solution schemes like ICE (Implicit Continuous Eulerian). Although they usually give . however. the conservative property is as important as the stability because the error from nonconservation may accumulate over many time steps.88 the domain boundary is equal to zero. the same exchange term should be used for neighboring meshes. When we are concerned with a steady state solution. For a long time transient.4.2 Physical Constraint The physical constraint is closely related with the stability in a practical sense. A scheme may be both stable and nonconservative and also both unstable and conservative because the conservative property is independent of the stability. the conservative property may not be crucial. + CI 4. Although a conservative form is preferred to a nonconservative one. where the convection terms of momentum equations are linearized in a nonconservative way. All the explicit. implicit and ADI schemes can be made conservative by using appropriate differencing forms for the convection term. In order to guarantee conservation.
new values of 4 are given as follows.4.2. )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.89 similar constraints on the time step size and mesh spacing. The Courant condition and Hirt's stability condition [36] that the finite domain of influence should at least include the continuum domain of influence are good examples of the physical constraint condition.1 Diffusion There is a maximum net flow that can occur between neighboring meshes by diffusion. they are from entirely different origins. 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.D .
63 is unconditionally stable. t' > Q' .64) Eq. 4. there will be a damping oscillation in the solution when the time step size is greater than that given by Eq. R AxAy(4) C) > 2D (R Ax CyAt 2 DAt X2 < 1  (4. the solution will be unstable or stable with damping oscillations. d 2D t < 1 (4. Both of the oscillation and instaThe ADI bility should be avoided in a transient problem.65 may be a too conservative criterion in a real problem. . When the physical constraint is ignored and the time step size greater than that given by Eq. 4.65 is used. should still hold.64 or 4.65) Ax Equation 4. Therefore the physical constraint should be respected in addition to the stability condition to get a meaningful transient solution. we have the following physical constraint.65. however.90 Since the relationship.64 may be extended to a ddimensional case as follows. 4. 4.64 or 4. scheme given in Eq.
66) When a time step size greater than the Courant limit is used.3 Accuracy Realistic interpretation of a finite difference equa tion comes from an integral form of the conservation .2. At < 'x u (4. It is based on the assumption that 4A is the repreIf At is greater sentative value of the control volume A.4.2 Convection C A u B C A ). Therefore. the damping oscillation or instability may occur as in the case of diffusion. Then uAttay is no more an appropriate expression for the convection quantity during At. 4.4. a portion of the control volume C will cross the interface S.91 4.B 4y S The explicit and donor cell treatment of the convection term gives the convection quantity at interface S as ulAAtAy. than the Courant limit.
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 . I I t i r _ I I [ n n+1 n n+l _ 0 It Explicit Implicit CrankNicholson $i+.
The accuracy expressed in terms of O(At). It depends on the maximum eigenvalue of the iteration matrix. Since the exact profile depends on the Peclet and Strouhal numbers of the governing equation. If its absolute magnitude is less than one. The infinitesimal approximation of differential terms is not appropriate in most problems of our consideration. the accuracy is usually the least important among the given four considerations. accuracy depends on how well we can compute the convection and diffusion quantities at control surfaces over the time At. 4. whether the error introduced at a certain step will increase or decrease as the calculation goes over to the following steps. appropriate schemes should be chosen according to those numbers or weighting should be done between neighboring time steps and mesh points to increase the accuracy of the solution.4. The Von Neumann .4 Stability The stability is a mathematical problem.93 equation. However. the accuracy depends on the profiles 4 in time and space because the exact profiles assumed for are not known a priori. In other words. the error propagation is suppressed and the largest error contribution will be from the previous time step. The stability can be checked by the theorems about matrix eigenvalues or Von Neumann analysis. 0(Ax 2 ) is The also not appropriate for finitely large At and Ax.
implicit and ADI schemes using the Von Neumann analysis.67) The solution 4. (4. but its algebra may get complicated.1.4.94 analysis is to expand the solution of a finite difference equation in Fourier series and check the decay of each mode separately.4. The stability is not a sufficient conIn the fol dition for an acceptable numerical scheme. is expanded in Fourier series as follows. 4. 4. n+l i n €i+i I + u i+1 n ii 2Ax i_1 lt n n i+l . 1 4n = E~nei where I = .4. stability criteria will be derived for the explicit.4.1 One Dimensional Central Differencing of Convection Term The finite difference equation is given for a one dimensional general conservation equation with central differencing of the convection term..2 2i + D i+1 i n 2 i i1 (4.68) .1 Explicit Scheme The Von Neumann stability analysis of a finite difference equation is simple in its idea. lowing sections.
69) Dbt Dt 2 Tx. Therefore. mode to achieve stability.72) In order to get the stability of Eq.70) Equation 4. the following expression for = lCxsin xI where cx d x uAt A tx  2d (lcos x) (4.95 The absolute value of should be less than one for every When Eq. in Eq.70 may be rewritten as follows by defining cos6 x as t.71) The function f(t) will be defined as follows.67. 2 11 = [1 .68 is substituted results.2dx(t) where t = cose 1 < t < 1 ] 2 +C 2(1t 2) 1 (4. 4. 4.67. the function . 4. f(t) = i 12 = (4d2_ 2 )t2 + (Sd 2+4d )t + C 2+4d 24d (4. [i .2dx(cosex) 2 + [Cxsinex 2 < 1 (4.
The parabola f(t) always meets the taxis because the discriminant is always greater than or 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.73) for any t in [1.96 f(t) should be less than or equal to zero for any t in [1. D(Discriminant) = (C 2 .74) One of the two meeting points with the taxis is the point t=l because f(l) is equal to zero.1].73 can be solved by a graphical method using the characteristics of a quadratic equation. f(1) = 0 (4. Therefore. f(t) < 0 (4.1] Equation 4.75) The parabola f(t) may have three distinct shapes according to the sign of the coefficient of the second order term. In case (1) f(t) is a parabola concave upward and f(l) should be less than or equal to zero as shown in Fig.10(1). .2d )2 > 0 (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. Three distinct shapes of the parabola.97 f(t) f (t) f (t) 'I I 4. 4.. r I I t i t 1 1 t 0. f(t).. according to the sign of the coefficient of second order term .10.
4.C (4.4. 4.76) In case (2) f(t) is a straight line and f(l) again should be less than or equal to zero. we get the following stability conditions.77) Summing up the results in Eq.77. t=l as shown in Fig. 4.10(3).dx) is shown in Fig. 4d 2 .1.11. 4. 4d 2 C 2 4d C > 0 < 0 : : 0 < d dx > < 2 Cx The stable region in the coordinate system (Cx. 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.76 and Eq.98 f(l) = 8dx(2dx1) < 0 0 < d < (4. 4.2 One Dimensional Donor Cell Differencing of Convection Term The following Eq.78 is a finite difference equation with donor cell differencing of the convection term.4. . 4.2d Axis of symmetry: t = 4dx dx > C 2 x > .
11.d ) for an explicit scheme with central differencing of convection term in a one dimensional problem . Stability ondition in the plane (Cx. 4.99 d x 1 1 0 1 dx = C2 x 2x C Fig.
100
n+l
i
n At
At i
n
+ u i Ax
n
ii
n
= D
i+l
n
i
n
ii
Ax
2
2
(4.78)
where the velocity u is positive. The Von Neumann analysis of Eq. 4.78 gives the following stability condition.
2
It
1(C x +2d [
x
)(1cose x ) 2 + [C sine ) x
1
(4.79) (479)
can be shown that the value of IC12 in Eq. 4.79 does not
change when the velocity u is less than zero, if the Courant number, Cx, is defined such that it is always positive. The function f(t) is defined as follows to make it easier to solve Eq. 4.79.
f(t)
=
12
1
S[(C x+2d)
2Cx
]2 t
+ [2(C x+2d )2 2(C +2d
) ]
+ 2(C x+2d )]t (4.80)
+ [(C +2d ) +C
The function f(t) in Eq. 4.80 should be less that or equal to zero for any t in [1,1] for the stability of Eq. 4.78. f(t) < 0 for any t in [1,1] (4.81)
The parabola f(t) meets the taxis at the point, t=l. f(1) = .0 (4.82)
101
Equation 4.81 canbe solved by the graphical method as in the previous section. 2 2
(C +2d )
 C
> 0
:
0 < Cx +2d X < 1 
(4.83)
(c x +2d x)
2
C
2
x
< 0
d
>(C 2C ) dx x
(4.84)
The results, Eq. 4.83 and Eq. 4.84, are shown graphically in Fig. 4.12.
4.4.4.1.3
Two and Three Dimensional Donor Cell Differencing of Convection Term
The two dimensional extension of Eq. 4.78 is given in the following. n+l ij At n D . i+D 3 24
Lx
n ii
n + u ij Ax n
n i1 n
n + vii by
n iji n
+ 4
A2
1il3
n n 4  24 + D2+l ,_2
ay
+
ii
(4.85)
The Von Neumann analysis is applied to the finite difference equation, Eq. 4.85, yielding the following result.
 1 = C x (le
lex)

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

Ax
C
 vat
d

DAt2 x2
x
Y
DAt Ay2
The Courant numbers, Cx and C y , are defined so that they are always positive, independent of the directions of the velocities u and v. Equation 4.86 can be simplified
by defining al and a 2 as given in the following Eq. 4.87. = 1 + al1 where a= 1 C x  2d x + (Cx+dx)e x I6
x +
a2
(4.87)
+ de x
le
x
S
=
2
C
Y
2d
+ (C +d )e YY Y
I6
x
+ de Y
Iy
The stability condition is that the absolute magnitude of 5 should be less than or equal to one for any values of 8x
e and By.
From Fig. 4.13 and Fig. 4.14 it can be seen that
indea1 and a2 should satisfy the following conditions pendently. a + 1
4
(4.88)
(4.89)
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 . 4.104 Im \ .13. .1 SRe 2 al +2 1+a +a 2 Fig.
. .Re Both al and a 2 should be in the shaded circle to ensure that Ial+a 2 +1I< i. The region in the complex plane where the amplificatior factor. 4..105 .14. Fig. should exist for the stability of an explicit scheme with donor convection term in cell differencing of a general two dimensional problem .
4. which depend on both the mesh spacing. the stable range of Cx should be determined in terms of the cell Reynolds number Cx/dx'. I 1 2 + (1+)e Rx Rx 2 + (1+)e a1 = Cx [1 + Ie 1 3 . At.15.4 Stability Condition in Terms of Cell Reynolds Number The stability conditions in the previous three sections are in terms of the dimensionless numbers. Ax. The results are given in Fig. 4.88 and Eq.1.88 and 4.e Rx .4. 4.e Y) (4. which is independent of the time step size. the stability However. 4.16.dx) and (C y. Therefore. At. When Ax and At are already given.91) . The stability condition is again given by Eq. Cx and d . The stable region shrinks on a linear scale with dimensionality.89 and al and a2 may be rewritten as follows. and time step. A three dimensional case can also be solved through the same procedures and the results are given in Fig.d) should be in the shaded region for stability. where the points (Cx.4.90) a 2 = C [1 y R R Y + R (4.106 Both Eq. our usual concern is can be checked directly. 4. 4.89 can be solved by the graphical method as in the previous sections. the stable time step size for a given mesh spacing.
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.dx) and (Cy. . 4.idition in the planes (Cx. Stability cc.
dx) etc. for an explicit scheme with donor cell differconvection term in a general three encing of dimensional problem . 4. Stability condition in the planes (Cx.108 d xy(d or d z ) Cx d dx + 1 1 Cx (C d! S3 2 .(C Cx x y or Cz) z d ) Fig.16.
f(t) = C (1+K) 2 (lt) .1] f(l) = 0 (4.88 and Eq.92) Equation 4.C (+2x .93) ) where 2 K = 1 +.95 ) The function f(t) satisfies Eq. 4.94 ) D(Discriminant) = C 2 (K2C ) 2 > 0 (4.R and x t = cosex The stability condition is now that f(t) should be less than or equal to zero for any t in [1. 4.C K(lt) + C ( l_1t = (C 22 KC 2 2 )t2 + (2C 2 22 22 K +C K)t + C 22 K  C K + C 2 < 0 (4.95.) RX (1cose )) x + [C sine] x x < (4. and has distinct shapes for the Following three cases. 4. 4.90.92 can be simplified as follows.94 and Eq. [ . (1) Cx 2 K2 1) > 0 concave upward .__LI___IIII__LI*X__ ID I 109 The following condition for Cx and Rx can be derived from Eq.
The stability condition in Fig. R > 0 x : 0 < C < x 2 2 2(1+) x and (4. although it can be relaxed if there is a constraint on the velocity direction and mesh spacings such that the following' relation is satisfied.17.97) The results in Eq. therefore the subscript x may be replaced with the subscript y in Fig. 4. 4.17 is applicable to all general 2D problems.17. (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 are shown in Fig.yu < 1 (4.98) Ax . t=l. 4.96) R 1 x < 0 < C x < (1+2) Rx . 4. V f . to zero.110 (2) C 2(K 21) = 0 linear (3) 22 C 2(K 2_1) < 0 concave downward For case (1) and (2).96 and Eq. The same result will be obtained for the ydirection. solution procedure is similar to those in the previous sections and the results are given in the following as. 4.
17.Cy) for an explicit scheme with donor cell ditferencing of convection term in a general two dimensional problem .111 x (C y C x  2 1 2 (1+ L) R x C = 2(1+ ) x 2 0 R (R x y Fig.Cx) and (Ry. Stability condition in the plane (Rx. 4.
18.112 1 1 x CX x 1 (l+f) (1+ x ) 1+f 1 1 f 0 x V where f =y u Ax Fig.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.
4.100) For a general two dimensional problem.17 and numerical experiments. 4. 4. (4. parison between the analytical result in Fig. 4.uLx.99) The stability condition is shown in Fig. and it can be seen that the implicit scheme is unconditionally stable if the diffusion constant is than . d >  greater Cx (4. 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 (4.99 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. the stability condition is found to be the following.C Y y .20. 4.Cx d > .19 and Fig.102) dx > .99) (4. 4. 4. 4.2 Implicit Scheme The following Eq.101) (4.17 should be Table 4.18.2 shows the com 1 l+f in Fig.113 Then the number replaced with on the C axis in Fig.4.
780 0.3 8.488 0.720 0.333 4.2.833 8.260 0.540 0.380 0.629 2.167 2.660 0.499 0.333 833.147 0.019 Numerical Experiments 0.333 83.403 0.660 0.120 0.114 Maximum Stable C. R x 0.060 Table 4.083 Von Neumann Analysi s 0.120 0.180 0. Comparison of the stability conditions in terms of the cell Reynolds number by Von Neumann analysis and numerical experiments in a two dimensional explicit scheme with convection donor cell differencing of term .
C x 2x x 2x C x dx 1(C2+C 2xx ) Fig. Stability cndition in the plane (Cx.19.dx) for an implicit scheme with donor cell differencing of convection term in a one dimensional problem . 4.115 d =  x d =.
20.Cx) for an implicit scheme with donor cell differencing convection term in a one dimensional of problem .116 Cx=1  2 Rx 2 1 0 R x R C x R +2 Fig. Stability condition in the plane (Rx. 4.
104) .63 gives the following stability condition. 4. One of the formulations has both the x and y direction components in each step.3 ADI Scheme There are two possible formulations of the Alternate Direction Implicit (ADI) scheme given in Eq.117 4.4. The Von Neumann analysis of the first formulation in Eq.63.+d )] C ) {[1+(lcosex +dx)) +[fsinex }{[l+(1cosy) (+dy) + Y[siney } (4. treating them implicitly one by one at each fractional step and the other formulation has only the x or y direction component treated implicitly in each step.jto be less than or equal to one.4. 4.103) jsj <1 In order for I. C d x>2 (4. 2 _C C 2 {[(cos )(+d X))] C [sine I2} {(1cos6 C 2 )(. the following condition should be satisfied.
C d x > 2 2x (4.107) Therefore the second formulation of the ADI scheme in Eq. 4. .63 has the same stability condition as the fully implicit scheme.118 The same condition should hold for the ydirection.105) The stability condition of the second formulation in Eq. 4.63 can also be given by the Von Neumann analysis as follows.106) C d y  >  ~ 2 (4. C d y>  2  4 (4.
1. energy convection and heat transfer models are considered and the resulting flow fields are given in Fig. 5.1 and Fig.2. For example. density is determined by temperaSince the natural convection of air ture and pressure. 5. 5. Finally. occurs as a result of the buoyancy force due to density . The physical models are tested in a natural convection problem and the numerical schemes are tested in a simple geometry to compare numerical solutions with exact solutions.1 Updating of the Reference State The state of air is determined by any two independent properties. 5. Updating of the reference state. Natural convection occurs due to the heat transfer between the obstacle and containment air.1 Natural Convection Results The containment is modelled as a two dimensional rectangular compartment with an obstacle as a heat sink.119 CHAPTER 5 RESULTS Some calculations are performed to validate the physical models and numerical schemes in the previous chapters. the ADI scheme is explored to use a time step size larger than the Courant limit to save the overall computational time.
5. 0. = 677. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 10.4 sec AT = 100F T .8 sec .29 Ibm A= 2 ft2 T.1.8537x10tu/ft20 F sec (Cp) = 1. = 680F air T = 580F Fig.0 Btu/lbm 0F (Q ob = 208.120 p ).5 ft/sec h = 1.
5 sec . 5. T ab = 580 F Fig. 5 ft/sec h = 1.29 ibm (K 2 A= 2 ft T.4 sec 1 0 AT = 10 F T air = 680F .0 Btu/lbm OF (c )b = 208. = 677.8537x10 tu/ft2 F sec p)ob = 1.2. Natural convection flow field due to heat transfer between the obstacle and air in the containment at t = 62.121 > 0.
Pold' Pnew: . temperature and pressure are equally important parameters in determining the flow field.cell where Pold' Pnew: Density of the fluid at old and new time steps. Told' Tnew: Temperature of the fluid at old and new time steps. not on the pressure because of a high modulus of elasticity. The density change of liquid. depends primarily on the temperature. c : Specific heat of the fluid at constant volume. The equation of state of an ideal gas is given in the following. time Pressure of the fluid at old and new steps.total Pold Told new old Qcell cv Mf. new Pnew RT new (5.1) T Pnew = old + Ttotal c v M f. however.122 gradients.
3 Heat Transfer Modelling The heat transfer to the obstacle is modelled as a natural convection from a vertical surface. the atomic gases have different values of k.4 for air..total : Mass of the fluid in one cell and in the total domain.4 when internal energy is used for enthalpy in air flow.1. .123 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.cell. diatomic and polyTherefore.1. An experimental correlation is given in Eq. convection energy transfer should be in terms of enthalpy instead of internal energy. The energy convection will be underestimated by a factor of 1. but they may also be based on the vertical height of the obstacle. 5.2 in terms of the dimensionless numbers. 5. k = c /C for liquid and 1. Qcell total : Heat input to the fluid in one cell and in the total domain.0 Monatomic. The Nussett and Grashof numbers here are based on the axial length of one computational mesh. Mf.2 Energy Convection Another difference between liquid and gas is the ratio of the Specific heats. 5. The value of k is 1.
h = 1.021 n 0.7134 x 108 Nu = 89.8537 x 10 .0763 ibm/ft 3 g = 32.25 0.sec .4 The following data are used to calculate the heat transfer coefficient at the obstacle wall. This may be a typical value for the natural convection heat transfer in the containment without any phase change.876 The heat transfer coefficient is given as.2) 3 2 Gr = gS(AT)L p 2 (5. .2 ft/sec' 5 V = 1. Gr = 9.923 x 103 Therefore.124 Nu = C(GrPr) n ( 5.109 > 109 C 0.2179 x 10 B = 1. L = 2 ft AT = 500 F p = 0.555 0.3) Pr = 0.708 (air) GrPr 105 .4 lbm/ftsec Btu/ft 2 F.
4) where CQ is the source term in the energy equation of fluid. The time constant. 5. CQ = Cpob (TT) (5. (Mc )ob (MC) obf hApo b ob f (5. Obstacle density. Specific heat at constant pressure. the heat . If the flow is turbulent.125 The VARR input [21) requires the time constant of the heat transfer instead of the heat transfer coefficient.4 sec. T = 677. Heat transfer coefficient between the fluid and obstacle. can be obtained from heat balance between the obstacle and fluid as follows.2. Heat transfer area between the obstacle and fluid over one computational mesh. pf Pob: c h : Fluid density. Therefore.1 and Fig.4 Turbulence Modelling Laminar flow is assumed in obtaining the results in Fig.1. 5. T.5) where M A : Mass of one computational mesh. 5.
to the obstacle will be enhanced and the diffusion transport of momentum and energy will also be enhanced in the containment air.2 Numerical Diffusion The truncation and crossflow diffusion errors are compared in a simple geometry where an analytical solution can be obtained.5 show that the numerical solutions is close to the analytical solution and the truncation error can be neglected. The results in Fig. 5.3.126 transfer. 5.6 and Fig. 5.4 and Fig.2. The corrective scheme is tested in recirculating flow problems for mesh point and mesh interface implementations.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.3(B) is skewed at 450 to the mesh orientation and both truncation error and crossflow diffusions are included in the solution error.7 show that excessive numerical diffusion has occurred due to the . Figure 5. The skew differencing and corrective Huh's formula schemes are tested for various problems. The flow in Fig. 5. 5. 5. 5. steadystate problem in Fig. is compared with De Vahl Davis and Mallinson's in predicting the magnitude of crossflow diffusion. 5. The flow in Fig.
0 BT S=0. Problem geometry for the two cases.0 u = v = 7.0 B T_ 0.0 u = 10. 5.127 (A) .0 x BT  @x a( aT + 2 2 a2 ay ) Th= 10.0. for evaluation of numerical diffusion .0 ay AB=BC=5.3.07 (B) =0.0 B T specified Fig.0 a= 1. 0 ST By =0.0 (A) (B) AB=BC=5. (A) parallel flow and (B) cross flow.
128 Analytic solution Numerical solution with Donor Cell T 10. spatial coordinate Fig.3(A) . Comparison of the analytic solution and numerical solution with donor cell differencing of convection term in 10x10 meshes along the line CD in Fig.4. 5. 5.
5.~ 129 Analytic solution Numerical solution with Donor Cell C spatial coordinate Fig.3(A) .^XI ill I~LX_L~LI. 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.__LL l.~ I~.~l r.5.__~ .
6.130 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Df cf T 10 spatial A coordinate Fig. 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. 5.3(B) . Comparison of the analytic solution. 5.
analytic solution with increased diffusion constant and numerical solution with donor cell differencing of convection term in 6x6 meshes along the line CA in Fig.7.131 Analytic solution Numerical solution with Donor Cell Analytic solution with D+Dcf T 10 / I / / . 5. .3(B) .J 0 I' C spatial A coordinate Fig.. Comparison cT the analytic solution.3(B) s' / line CA in . 5./ Fig. 1/ 5.
It can be seen that the former result is much better than the latter.8 has the arbitrary angles of 6 and 61 such that 6 = 600 and 61 76.10 by De Vahl Davis and Mallinson's for the problem in Fig. 5.2 Validation of Huh's Correction Formula The crossflow diffusion can be predicted by Huh's formula. Figure 5.54. Eq. 10 x 10 meshes and Fig.3 Skew Differencing Scheme Two skew differencing schemes. 5. The problem geometry in Fig. 5.7 are for 6 x 6 It is seen that a finer mesh gives a better Since the crossflow diffusion constant is pro portional to the mesh spacing. 5. 4. De Vahl Davis and Mallinson's and Huh's formulas give the same crossflow diffusion constants for the case.5 and Fig.2. Figure 5.132 crossflow diffusion. 5. 5. solution. meshes.11 and Fig. more diffusion has occurred in Fig. 5. 5.9 is the result by Huh's formula and Fig.6 are for 5. are tested in Fig.53 and Eq.8. 5.4 and Fig. 5.6 and Fig. 5. Since the corrective scheme is based on the validity of those correction formulas.6. Raithby's and Huh's. 5.810. 4.7 show that the numerical solution can be reproduced by the analytical solution with increased diffusion constants by the amount given by Huh's formula. it is necessary to test them for various cases. 6 = 6 = 450. Figure 5.7 than in Fig.12 for a simple pure .2.
5.8.0 v = 8.6603 DDM= 0.335 Huh A e1 = 76.133 Yo 2 T=0 D a = 1.0 Ax = 0. Problem geomctry with arbitrary angles of 6 and 81 to compare De Vahl Davis and Mallinson's and Huh's formulas for prediction of crossflow diffusion constant .0267 u = 5.810 8 = 600 8 Meshes 24 Meshes Fig.7674 DH u h = 3.2406 by = 1.
by Huh's formula along the line BD in Fig. 5 0 spatial coordinate Fig. Comparision of the numerical solution with donor convection term and cell differencing of analytic solution with increased diffusion constant.134 Analytic solution with D+Dcf Numerical solution with Donor Cell T 10. (D+Dcf).9. 5.8 . 5.
10. 5. solution (D+DDM).135 Analytic solution with D+DDM Numerical solution with Donor Cell T 101 spatial coordinate Fig. 5.8 . Comparision of the numerical solution donor cell convection term and analytic differencing of with increased diffusion constant. by De Vahl Davis and Mallinson's formula along the line BD in Fig.
75 81.5 0 6.11. Comparison of the true solution and solutions by donor cell scheme.75 25 U 12.25 18.25 65.75 75 50 U Skew differencing by Raithby 50 100 100 100 100 50 100 100 100 50 0 100 100 50 0 0 100 50 0 0 0 50 0 0 0 0 50 34.25 0 Skew differencing by Huh (True solution) 100 100 100 100 50 0 0 0. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure convection problem with e = 45* .38 50 31.63 87.136 Donor cell 100 100 100 100 100 93. 5. 0 100 100 100 100 100 50 100 50 0 50 0 0 V Fig.5 68.
5 0 0 63.99 10.6.26 80.01 55. 5.3E 21.43 7.2! 90.50 2.74 38.8922.25 53.52 1.11 0 11.33 70.137 Donor cell 100 100 100 100 50 .67 5.5E 25.12.E 21 0 12 True solution 3.59.96 17.71 20.85 0.6 0 Skew differencing by Huh 100 90.88 81.56 1.3: 11.91 71.22 9.9 33.11 3.91 31.430 0 50 0 Comparison of the true solution and solutions by donor cell scheme. skew differencing scheme by Raithby and skew differencing scheme by Huh in a pure covection problem with e = 63.25 0 0 0 .3' 40.13 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 6.86 16.A Skew differencing by Raithby 100 100 100 100 96.430 . 62.83 72.70 4.1 6.25 37.5C 100 100 50 Fig.25 1.47 23.
former is conservative while the latter is not. The details of those schemes are Figure 5. Raithby's scheme is an Eulerian Therefore.1 and 4.2. the However. but less than inthe donor cell scheme. however. 8 = 1e= 45*. The purpose of the corrective scheme is.11 has a flow given in sections 4. 5.13 and Fig.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.2. 5. both of them are conservative in a unidirectional flow.138 convection problem. 5.3. method while Huh's scheme is Lagrangian.14 show that the analytical solution can be reproduced by Huh's skew differencing scheme for the case.430 angle and neither Raithby's nor Huh's scheme can reproduce the true solution.3. 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. to obtain the true solution by subtracting the additional crossflow diffusion . Figure 5. It is indicated that Raithby's scheme may give unphysical results in problems with steep gradients of the quantity under consideration. Some crossflow diffusion has occurred in Huh's scheme.12 has a flow at a 63. Fig.
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.13. 5.3(B) . 5.
14. 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) .140 Analytic solution Numerical solution with Donor Cell Skew differencing scheme by Huh 10 5 0 spatial A coordinate Fig. 5.
430. 5. Figure 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.3(B).17 and Fig. are introduced in Chapter 4.19 is a hypothetical 3 x 3 recirculating flow field where the inlet boundary values are specified on the left and bottom surfaces.141 constant from the total diffusion constant of the finite difference equation. Figure. 5. 5. The corrective scheme is also tested in the problem geometry of Fig. Figure 5.16 has a flow with 8 = 63. 5. donor cell solution and the solutions by mesh point aand mesh interface implementations.15 numerical solutions are given for a simple pure convection problem with various diffusion corrections. The corrective scheme gives a physically reasonable solution. 5. Two implementation strategies of the corrective scheme. Both of them are tested in a recirculating flow problem with pure convection.21 gives the true solution.20 shows the diffusion corrections of the mesh point and mesh interface implementations at each interface. In Fig. although not identical to the true solution. The flow is at a 45* angle with respect to the mesh orientation. Fig. mesh point and mesh interface. 5. The mesh point implementation gives an unphysical solution . Fig. The full correction of the crossflow diffusion gives the true solution without any numerical diffusion error.
1 10 8. 5.08 5.15.0 2.5  3.41 u=v=l Ax=Ay=l D =D =0.5 (True Solution) 9.0 10 7.13 7.71 0.88 5 4.12 5.01 7.142 D=0.3 D=0.5 0 1.59 7.92 5. 0 V ________ D=0. Donor cell solutions with various diffusion corrections for a pure convection problem with e = 450 .0 2.5 x y Fig.88 10 5 0 2.75 6.0 2.99 D=0.0 8.71 5.29 0.0 1.25 0 5.29 5.
07 10 3.430 .56  2.07 0.41 0.143 D=0. 5.37 3.04 4.11 True Solution V 63.93 10 5.91 0.73 10 7.430 u=l.11 0.07 2.11 6.96 3.10 9.16.33 1. v=2 x= y=l 2 D =D = x y 3 Fig.59 1.82 0. 0 D2 3 0. True solution and donor cell solutions with and without diffusion correction for a pure convection problem with e = 63.
17. Comparison of the analytic solution. numerical solution by donor cell scheme and numerical solution by corrective scheme for 10x10 meshes along the line CA in Fig. 5.144 Analytic solution Numerical solution with donor cell Solution by corrective scheme 10 A spatial coordinate Fig.3(B) . 5.
3(B) . 5. numerical solution by donor cell scheme and numerical solution by corrective scheme for 6x6 meshes along the line CA in Fig.145 Analytic solution Numerical solution with donor Solution by corrective scheme 10 5 0 spatial coordinate Fig. Comparisoin of the analytic solution.18. 5.
5. 3 ) 2 3 2 Lx = Ay = 1 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 .19.
2 0 0 0 mesh interface implementation Fig.19 .75 0 0 0.147 0.20.75 1.125 0.2 0.25 mesh point implementation 0. 5.5 0 0 5 0 0 .75 1.25 1.5 0. 5.75 0.625 0.625 0 . Diffusion coastant corrections at each interface for the mesh point and mesh interface implementations of the corrective for the recirculating flow field given in Fig.75 1.
26 8.75 6.21.08 9.86 12.82 8.21 7. 5.83 6.148 True solution Donor cell mesh point implementation 5.51 10 5 0 Fig.19 . 5.16 11.10 9.10 8.18 7. 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.16 6.94 9. Comparison of the true solution.21 mesh interface implementation 0.
The steam concentration distribution at the beginning is given in Fig. The true solution cannot be reproduced because some of the information about the flow field has already been lost in Fig. 5. Another test calculation is done for a two dimensional. Figure 5. 5.25 and Fig. the steam concentration for the new time step can be calculated in an explicit way. Since the steam concentration distribution and flow field are given at the start of the calculation. 5.28 is not a reasonable solution while the mesh interface implementation result in Fig. The result in Fig.22 shows the flow field in the containment and more detailed information about the flow field is given in Fig.28 show the steam concentration distributions after one time step. 5. flow field is used here to test the mesh point and mesh interface implementations of the corrective scheme. Figure 5. 5. rectangular containment with an arbitrary recirculating flow field.19.3 sec with the corrections in Fig.25 and Fig. After a while a flow field is set up in the This containment by the input mass and momentum of steam.27 and Fig.149 while the mesh interface implementation gives a physically reasonable solution.24 in which the source mesh has the highest steam concentration. 0.2 kg/sec. 5. respectively.23.27 is physically reasonable and even better . 5.26. There was initially air in the containment and steam is introduced in the source mesh at the rate of 0. 5. 5. 5.26 show the diffusion constant corrections of the mesh point and mesh interface implementations and Fig.
5.22.5m Ay = 1. 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 .0m Fig.150 Ax = 0.
3074 0.3 095 1.0348 0.4870 0.2216 0.2460 0. 1317 0. 3563 0.1276 0 .531 0. 5.1422 2148 0.7126 0.2159 0.2474 04 1379 0.207 2.1385 0. 0735 0.5174 1.3957 0.2435 0.2190 0. 861 1.456 0.4920 0 . 5955 SOURCE 0.3253 0.151 0.93 2.5 m Ay = 1 m Fig.5893 0. 0648 1.34 0.1139 0 .0937 0 1.0492 0 .2629 0.23. 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 .1229 0.0127 0. 2214 velocity [m/sec] Ax = 0. 045 0.
1266 0.1085 0.152 0.1244 0. 5.0011 0.0298 0.1750 0.0346 0.0002 0.0179 0. Initial steam concentration distribution in the two dimensional containment to test the mesh point and mesh interface implementations of the corrective scheme .1712 0.1420 0.1178 0.1710 0.0305 0.0676 0.5m by = Im Source: steam concentration in Ikg/m 3 0.0031 Ax = 0.24.0620 0.0049 0.1592 0.0819 0.0095 0.2 kg/sec of steam Fig.0007 0.0358 0.0136 0.1859 SOURCE 0.
0 278 0.25.1452 0.3262 0.153 0. 2422 0.2740 0.0 183 Ax = 0.0193 0.0937 0.0600 0. 0152 0. 5.0440 0. 0.1701 0. Diffusion co.C209 0.1 389 0.0348 0. 0227 0.1014 258 0.1451 0. 0.0300 0.0224 0.0255 SOURCE 0.1694 3106 0.. 0396 0.0311 0.0311 0.0253 0.1583 0.0590 0. 0488 0.0 362 0.1279 0.0794 0. 0086 0.5m by = Im Diffusion constant in Im2/sec] Fig.stant corrections of the mesh point implementation of the corrective scheme for calculation of the steam concentration distribution in the two dimensional containment . 0.0731 116 0.1763 0.1178 0.1854 0.
4732 0.0 111 0 Lx = 0. 0706 0. 0579 0.0287 0.0 923 0 0.0 257 0.5m by = lm Diffusion constant in Im2/sec] Fig.0443 0. 0240 0.0632 0.2529 0.2317 0 0.0192 0 0.0534 0.0 76 0.0518 0 0 0.26.9769 0 0 0.3694 0.2134 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 .1026 0 SOURCE 0 0 0. 5.0479 0.0215 0.154 0 0 0.0860 0.1149 0 0 0.
0448 0.0342 0.3 sec.0947 0.1243 0.0735 0.0046 Lx = 0.0016 0. 0.1363 0.0003 0.0195 0.1135 0.1733 0.0011 0.27.0465 0.1711 0.1847 SOURCE 0.0233 0.2 kg/sec of steam Fig.1475 0.0068 0.155 0. with donor cell differencing convection term without any diffusion of correction . 5.1321 0.0107 0.1623 0. Steam concentration distribution after one time step.1706 0.5m Ly = Im steam concentration in [kg/m 3 ] Source: o.0342 0.0833 0.
0355 0.0433 0.1493 0.0042 Lx = 0.1240 0.0709 0. 0.0465 0. with mesh point implementation of the corrective scheme .156 0.3 sec.5m Ly = Im steam concentration in Ikg/m 0.2 kg/sec of steam Source: 3 Fig.1747 0.1729 0.28.1911 0.1334 0.1797 0.0092 0.1646 0.0937 0.0005 0. 5.1247 0.0250 0.0029 0.0068 0.0287 0.2086 SOURCE 0.0010 0.0229 0. Steam concentration distribution after one time step.0815 0.
1628 0.1310 0.0360 0.0481 0.1112 0.1342 0. 5.1487 0. 0.1243 0.0947 0.0010 0.0192 0.0228 0.0058 0.1847 SOURCE 0.1738 0.0295 0.1794 0.0100 0.0039 ax = 0.1756 0.157 0.0424 0.2 kg/sec of steam Fig. with mesh interface implementation of the corrective scheme .5m by = Im steam concentration in [kg/m 3 Source: 0.0722 0.3 sec.0815 0.0003 0.29.0007 0. Steam concercration distribution after one time step.
32 shows the profiles of 4 at the same loca tion for the time step size of 1 sec.0 and the inlet boundary value of 4 is. Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable.158 than the donor cell solution. Figure 5. 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. 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. The results of the ADI and explicit schemes are compared in the problem geometry of Fig. In Fig. =10.31 shows the profiles of 4 along the dotted line for the time step size of 0. There is a source of 4 on the right boundary over two computational meshes and the profile of Q along the dotted line is affected by that source. 5.3 ADI Solution The ADI scheme is tested to use a time step size larger than the Courant limit. In the meshes at the top of the source mesh. which is exactly the .5 sec. 5. Figure 5.30 there is an inflow from the bottom surface at the velocity of v=1. 5.0. The ADI and explicit schemes give almost identical results during all the transient. when the Courant limit is 1 sec. it has the maximum time step size that can give a physically reasonable solution.30.
0 on this line are compared.0 Ax = Ly = 1.0 c= 1. v = 1.YXI~^. 5.30.0 Fig. Problem geometry for comparison of the explicit and ADI solution schemes .1I I _~ 1_.~ ~ILI_~_~LI*L~l*IIIU*XI 1 9 IB Source of ' 'A Profiles of 4 = 10.
0 coordinate 11.0 11.0 12.5 .0 12.01 12.0 11.30 the profile of sec and the Courant limit of 1.160 12. Comparison of the explicit and ADI solutions for along the line AB in Fig. 5.0 11.0 sec for Lt=0.0A 12.0 10.0 10.0 11.0 11. 0A Explicit ADI 10.31.0 10. 5.0 0.0 A iatlai B 10.0 10.0 Fig.5 sec 12.
0 sec and the Courant limit of 1. 5.0 16 sec 20 sec 12.0 4 sec 11.0 10.0 11.0 10.32.0 12.0 Fig.0 . Comparison of the explicit and ADI solutions for the profile of 4 along the line AB in Fig.0 12.0 11. 11. 5.161 1 sec 12.0 11.0 10. 12.30 for At=1.0 ~ spatial coordinate 12.0 sec .0 Explicit ADI 10.0 10.
Fig. 5.162 the Courant limit.31. 5.33 shows that the ADI scheme can use a time step size up to five times the Courant limit. No crossflow diffusion corrections are involved in the results of Fig.33. 5. Fig. The deviation for the time step size of 5 sec is expected to damp out as the calculation goes over to the next time steps because of the unconditional stability. there is some deviation between the two results. . Initially.32 and Fig. but it damps out rapidly and the ADI scheme gives almost identical results with the explicit scheme. 5.
33.0 sec along the line AB in Fig.5 sec ec 11. 5. 5.163 12.0 t spatial coordinate B Fig.0 At =0.30 with different time step sizes . Comparison of the ADI solutions for the profile of at t=10.
or superheated steam as may be appropriate. The total diffusion con stant is the sum of the laminar and turbulent diffusion constants. in the finite difference donor cell treatment of convection. truncation error and crossflow diffusion. The mass diffusion. Crossflow diffusion occurs due to the donor cell treatment of the convection term in a multidimensional problem. and turbulence equations are solved separately using that flow field. The thermal equilibrium is assumed and the phase change occurs to maintain 100% relative humidity.164 CHAPTER 6 CONCLUSION 6. energy. The .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 con tinuity/momentum equations are decoupled from the scalar transport equations and solved by the SMAC scheme to obtain the flow field.2 Numerical Schemes (1) There are two numerical diffusion sources.
Therefore. The corrective scheme is based on the fact that the additional crossflow diffusion can be predicted theoretically for every mesh at every time step. convection dominant. have been tried in order to eliminate the numerical diffusion. . The inclusion of corner points complicates the matrix structure and a fully implicit scheme should be used for maintenance of stability. It gives unphysical results for most recirculating flow and coarse mesh problems. The truncation error diffusion occurs in the flow direction while the crossflow diffusion occurs in the diagonal direction of neighboring mesh points.165 effective diffusion constants of the truncation error and crossflow diffusion are of the same order of magnitude. 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. (2) Two types of approaches. the corrective scheme is generally preferred to the skew differencing scheme. but not always. skew differencing and corrective schemes. Therefore. ~unx. It is conservative and an explicit scheme can be used without affecting the simple solution structure. The skew differencing scheme gives good results for some problems. most of the numerical diffusion error in a multidimensional. recirculating flow problem is due to the crossflow diffusion.
Although the Von Neumann analysis shows that the ADI scheme is unconditionally stable.2 imposes a maximum time step size that can give a physically . The results of numerical experiments are found to be consistent with the Von Neumann analysis. First. diffusion and cell Reynolds numbers. Implicit (ADI) scheme can be used to increase the time step size and decrease the overall computational effort.. are tried for the corrective scheme. The graphical method is used to obtain the stability condition in terms of the Courant. The mesh interface implementation is always preferred to the mesh point implementation in a recirculating flow problem. Cx. it has its own limitations due to the following factors. dx and Rx . the physical constraint in section 4. i. mesh point and They mesh interface. (4) The Von Neumann stability analysis is applied to various finite difference forms of a general conservation equation.4. are identical in a unidirectional flow.e. (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. using the characteristics of a quadratic equation.166 (3) Two implementation strategies.
Second. . the ADI scheme is a fractional step method. a steadyThe state solution may never be reached due to the fact that one time step is composed of a few fractional steps. For example. asymmetry of the given problem may give different results according to the sweeping sequence.167 reasonable solution.
The modified kE model in a low Reynolds number flow may be a guide in this approach.. to eliminate the crossflow diffusion. distance from the wall). transitional and laminar flow regimes. etc. The important parameters may be the velocity.2 Numerical Schemes There have been two approaches. it is necessary to predict the flow regime in order to use appropriate physical models.168 CHAPTER 7 RECOMMENDATIONS FOR FUTURE WORK 7. multidimensional recirculating flow problem. skewdifferencing and corrective schemes. The model should show proper limiting behaviors as the Reynolds number goes to infinity (turbulent) and zero (laminar). They may still be improved further to get an accurate solution in a transient. velocity gradient. previous time step information.1 Physical Models Since both the laminar and turbulent flow regimes are possible in a complicated geometry and flow field. geometry (e. A criterion for the flow regime should be developed in such a form that it may be implemented into a computer code. . The flow regime criterion may be replaced with a model that covers all the turbulent. time step size.g. 7.
169 Corrective scheme The corrective scheme has been successful by the mesh interface implementation strategy in the scope of this work. . 2. It may be possible to develop a reasonably simple form that may or may not include the corner points. The solution in Appendix A may be a useful guide in this work. The conservative form of the skew differencing scheme is complicated and time consuming in comparison with the corrective scheme. More tests and validation calculations are required in recirculating flow problems. Skew differencing scheme 1. ADI and implicit schemes although only the implicit scheme has been used up to now. The stability of the skewdifferencing scheme is open to question in the explicit.
" MIT Ph." Numerical Heat Transfer 3.R. H. Baliga. Briley. F. Bradshaw. T. Stewart. Phys.. Amsden. "The SMAC method: A numerical technique for calculating incompressible flows. and Harlow.D. E. W." Comp... John Wiley & Sons (1960).." NUREG/CR1148 (EGG208). and Harlow. 6.B. B.User's Manual. 94110 Amsden. D. 6.R... G. "Solution of the Multidimensional Compressible NavierStokes Equations by a Generalized Implicit Method. and Eng." Journal of (1968)." Journal of Comp. and Harlow. and Lightfoot. 34. 5473 (1980). "Transient Effects in Turbulence Modelling. F. and McDonald.R.A.H. D. "A New Finite Element Formulation for ConvectionDiffusion Problems.R. and Whitelaw. et al. Academic Press (1981). Engineering Calculation Methods for Turbulent Flow. Boyle. Broadus. F.R." HEDLSA2682 (October 1982). C..N.V.H. J.R. Phys. S." LA4370 (May 1970). W.. Briggs.A. Methods in Applied Mech.. Bird.H. Phys. Phys. [2] [3) [4] [5) [6] [7] [8) [9) [10) [11) [12] . W. A.G.H. "A Simplified MAC Technique for Incompressible Fluid Flow Calculations. "BEACON/MOD 3: A Computer Program for Thermalhydraulic Analysis of Nuclear Reactor Containments . "Hydrogen Distribution in a Containment With a HighVelocity HydrogenSteam Source.. A.170 REFERENCES [1] Amsden. R. "A Finite Difference Scheme for the Incompressible AdvectionDiffusion Equation." Journal of Comp.A. (April 1980). Briley.E. 332325 (1970).. H. Transport Phenomena. Bloom. 372397 (1977). thesis (1979).. and McDonald. et al.. 3. 393409 (1980). 24. and Patankar. "On the Structure and Use of Linearized Block Implicit Schemes. 233241 (1975). Comp. "Transport of Turbulence in Numerical Fluid Dynamics. A. Cebeci. P." Journal of Comp..
W. Varadi. thesis (1981). Clark." Comp." Computers and Fluids. and Nakayama. and Mallison. and Eng. L.. Cheng.H. S. 471484 (1967).L. J. and Design 69. G. Crowley.A Computer Program for Calculating Timedependent Turbulent Fluid Flows with Slight Density Variation. 1344 (1981). Academic Press (1964). K.G." MIT Ph.F. and Lewis. Phys. "SOLAICE: A Numerical Solution Algorithm for Transient Compressible Fluid Flows. and Ramshaw." Journal of Comp. "Studies of Numerical Methods for the Plane NavierStokes Equations." Journal of Comp.. 2943 (1976)... 7179 (1979).. J. of Fluids Supplement 2 (1969). Jr. "VARR2 .D. Y. H. S. De Vahl Davis. "Accuracy of Difference Formulation of NavierStokes Equations. B. J.K." MIT Ph. Chang. 95107 (1982). "Numerical Methods of Reactor Analysis.. P. T.W. P. 27. 1976).I." The Phys. Keller." Nuclear Eng. Cook.J. N. Methods in Applied Mech. and Romero... J... Hirsh.D. L.I. "SecondOrder Numerical Advection.D. S. Method for Convection in Numerical Fluid Dynamics.S..B..D. 1.. Vol. Hirt. C. M. "Coolant Mixing in the LMFBR Outlet Plenum. thesis (1980). Chang. Williams..H. "Simulation of Containment Cooling with Outside Spray After a Core Meltdown." Nuclear Eng.D." LA6236 (July 1976).. 32. "Buoyancy Effects in Mathematical and Numerical Turbulence Modelling. Covelli.B. G. Cloutman. Cebeci. "An Evaluation of Upwind and Central Difference Approximations by a Study of Recircul ing Flow. 1 (Nov.C...P. and Design 69. "Tensor Viscosity Dukowicz.." CRBRPARD0106. Vol. "Comparative Analysis of Numerical Methods for the Solution of NavierSTokes Equations. and Hansen.S. thesis (1977).. 127137 (1982). G. M. Nielsen. Phys..171 [13) Carbajo. [14] [15) [16] [17] 118] [19] [20) [21] [22] 123] [24] [25) . Chen." MIT M. 4. "Study of Nonhomogeneous Conditions in Reactor Containments Due to Separation Effects. R.
C. J.O. B. "Numerical Calculation of Multiphase Fluid Flow. T. Gupta.." Journal of Comp. Hirschfelder. 8195 (1981).F. J.H.W. 8. of Fluids Supplement 2 (1969). 1952 (1975). Vol. and Bird..B. Phys.A. A. of Fluids. [27] [28] [29) [30) [31] [32) [33) [34] [35] [36) [37) Hirt. R. R. C. Harlow. Molecular Theory of Gases and Liquids. Wiley.." Int. "A Comparison of Hybrid and Quadratic Upstream Differencing in High Reynolds Number Elliptic Flows. 101. 309312 (March 1975). McGrawHill Book Company. B. Harlow. J. F. Curtiss. 8093 (1968).." MITEL 81020 (June 1981). "Improved Form of the Low Reynolds Number kc Turbulence Model. Hirt." The Phys.. "Buoyancy Induced Fluid Motions Characteristic of Applications in Technology . C." The Phys.. "Computer Studies of Timedependent Turbulent Flows. "A Numerical Fluid Dynamics Calculation Method for All Flow Speeds. (1975). 29. F.. of Fluids Supplement 2 (1969). "Practical Investigation of Convective Difference Approximations of Reduced Dispersion. Harlow. and Amsden.O. and Golay. [38) ..H. Phys." Comp.172 [26] Fromm. Gebhart. and Amsden." Journal of Comp.H.P. "Heuristic Stability Theory for Finite Difference Equations..... 2. for Num.. Han.. New York (1954).A. 3. Hamza.. Methods in Applied Mech. Hinze... "Behavior of Buoyant Moist Plumes in Turbulent Atmospheres.M. C.H. M. and Marohar. Hoffman. and Eng. Vol. F. 528 (March 1979)." Journal of Comp. Vol. M. 197213 (1971). Phys. "Numerical Calculation of Almost Incompressible Flow. Methods in Eng.A. J.. "On the Use of Central Difference Scheme for NavierStokes Equations. 15. and Amsden.. 557573 (1980)." The Phys. J. 339355 (1968)..W.E." Journal of Fluids Eng..." Journal of Comp. Turbulence. Humphrey. G. Phys.W.The 1978 Freeman Scholar Lecture. A. and Launder.E.A. R. 18. A. 2nd ed. 17.
Heat Mass Transfer... J." Int. Manno..E. Frankfurt. Reece. and Bremhorst..J. V. 3. "Practical Evaluation of Three Finite Difference Schemes for the Computation of SteadyState Recirculating Flows. 68.G..A..P." MIT Sc. M.. Vol. C. M. Phys. Keenan.3633... and Spalding. 537566 (1975). W...V.P. V. [40) [41) [42) [43] [44) [45] [46] [47] Leschziner. "Analytical Modelling of Hydrogen Transport in Reactor Containments.D. Vol. F. J. "ADI Solution of Free Convection in a Variable Viscosity Fluid. and Eng." BF63.K. Steam Tables. B. M. translated (October 1980).." Report submitted to 2nd International Topical Meeting on Nuclear Reactor Thermalhydraulics (March 1982). 456460 (Sept. and Keyes. Launder. G. Vol.E. 148) [49) . 293312 (1980). B. "Progress in the Development of a Reynolds Stress Turbulence Closure. B.. Jr. 1981). K.. and Eng. 301314 (1972). J. and Launder. 221239 (1974). W.B." Comp." Journal of Comp. thesis (1983). "The Numerical Comparison of Turbulent Flows. "Analytical Modelling of Hydrogen Transport In Reactor Containments. Methods in Applied Mech. "Experimental Investigation of Hydrogen Distribution in a Containment of a Light Water Reactor Following a Coolant Loss Accident." Comp. and Launder.G. "The Calculation of Low Reynolds Number Phenomena with a Twoequation Model of Turbulence. part 3." Int.P. Fluid Mech. and Golay. Manno. 16. Methods in Applied Mech.. and De Braemaecker. Jones. 103. 16. D.H.. of the ASME. 269289 (1974). Lam. et al. G.E. Cl.173 [39) Houston.P. Langer. 23." Trans. Vol. Launder. B. "The Prediction of Laminarization with a Twoequation Model of Turbulence. 15. Heat Mass Transfer. "A Modified Form of the kc Model for Predicting Wall Turbulence. 11191130 (1973). John Wiley & Sons (1978). and Rodi.E..H.W. Battelle Institute e." J. Jones. W. J.
D. 75103 (1976). Hemisphere Publishing Corporation (1980). Roberts. 4. "A Note on the Empirical Constant in the KolmogorovPrandtl EddyViscosity Expression." Comp. 386389 (Sept.D. S.B.K. of the ASME.. The Properties of Gases and Liquids." The Phys. J. 17871806 (1972). "Convective Difference Schemes.. Heat Mass Transfer. Raithby.D. Runchal. Roache. Methods in Eng.E. Phys. "A Critical Evaluation of Upstream Differencing Applied to Problems Involving Fluid Flow.. 15. N. "On Artificial Viscosity.V. and Weiss. and Torrance. New York (1958).." Computers and Fluids. K. John Wiley' & Sons. Roache. W. P.. S.B. W. Computational Methods in Engineering and Science. 2.. of Fluids Supplement 2 (1969). Patankar." Comp. S. 1975). 20 (1966). Rodi. W. 9.. Computational Fluid Dynamics. G. Raithby. [52) [53) [54) [55) [56) [57) [58) [59] [60) [61) [62] [63) .. Methods in Applied Mech. Methods in Applied Mech.. 169184 (1972).K.. G.. Patankar. "Upstreamweighted Differencing Schemes and Their Application to Elliptic Problems Involving Fluid Flow.. Schonauer. "SkewUpstream Differencing Schemes for Problems Involving Fluid Flow.V.. Inc. "Numerical Experiments with a Difference Model for the NavierStokes Equations (Turbulence Model).J. Reid. Numerical Heat Transfer and Fluid Flow. 9.." Trans. Hermosa Publishers. (1977).D." Mathematics of Computation. Albuquerque.174 [50] [51) Nakamura. J. and Eng. T... Mass and Momentum Transfer in Three Dimensional Parabolic Flows. 10. R. K. "A Calculational Procedure for Heat. for Num. 191206 (1974). and Sherwood. McGrawHill. Pergamon Press (1982).J. New Mexico (1976). and'Eng." Journal of Comp. 153164 (1976)." Int. Turbulent Buoyant Jets and Plumes.C. Raithby. G.O. 541550 (1972). and Spalding. Vol... P. "Convergence and Accuracy of Three Finite Difference Schemes for a Two Dimensional Conduction and Convection Problem. Rodi. A." Int.
D. R. G. G. A First Course in Tur [65] [66) [67] [68) [69) [70] [71] Torrance." FATE80114 [73] (Jan.A. 411428 (1980).... 95102 (1950). No. Stubley.. Vol. Raithby.. 35.L.S.. 3. . J. 1981). MIT Press (1970). and Valentin. and Woolner.B.D. "Diffusional Properties of Multicomponent Gases. Y. C. Stuhmiller. Smith. Budden.. MatheResearch matical Sciences. Raithby. 1974). and Strong.E. "Numerical Analysis of Fluid Hammer Waves by the Method of Characteristics.H. Stubley.." Comp. G. A.W.R. bulence. 20. K. G. No. "A Novel Finite Difference Formulation for Differential Expressions Involving Both First and Second Derivatives. and Eng." Chemical Engineering Progress. J." Numerical Heat Transfer.. R. Methods in Eng.D. and Hutton.J. 1968). For Num. "Simulation of Convection and Diffusion Processes by Standard Finite Difference Schemes and by Influence Schemes.B.. Spalding. "Proposal for a New Discrete Method Based on an Assessment of Discretization Errors. Vol. Wilke." Int. 4." SAI74509LJ.. H. Trent. "Development and Validation of a TwoVariable Turbulence Model. Tennekes.B. 153168 (1982).D.175 [64] Shin.Dec.M." of the National Bureau of Science . A. Phys. A. 551559 (1972). K. "The Numerical Treatment of Advection: A Performance Comparison of Current Methods." Numerical Heat Transfer. "Comparison of Finite Difference Journal of Computations of Natural Convection.L. 46. Methods in Applied Mech. Strong.. and Eyler. 411428 (1980).G..D. 4 (Oct.. 2. L. J. and Lumley.B...." Journal of Comp. (Jan. 72B. 3. M. 220237 (1976).A. "TEMPEST: A Threedimensional Timedependent Computer Program [72] for Hydrothermal Analysis. D.
1 is2 reduced to the following form. 5.2 is simplified further by substitution of 4 for T as follows. Eq.1. A.3) ax . The boundary conditions are specified in Fig. Therefore.1 convection occurs in the x direction and diffusion occurs in both x and y directions. as follows. A.176 APPENDIX A ANALYTICAL SOLUTION FOR THE PROBLEM IN FIGURE 5.3(B) A steady state two dimensional energy conservation equation without any source is given in the follbwing Eq.1) In Eq.3(B) and the constants can be grouped into one constant.a2T aT 2 + a2 2 2c ax ax 2 y ay (A. 2 x 2 a2 ay = c 2 (A. a= k PCp C =2 u 2A.2) Equation A. pc u aT = k(a2T + 2aT 2 x2 p x ax ay (A. c..
a =  (A.7) d Y c .y) = X(x)Y(y) (A. I d2X + 1 d 2Y 1 = c2 C (A. The variable 1 is separated into two vari X and Y. Equation A.3 can be solved by the method of separation of variables. ables.8) dy where a > c > 0 . A.3 is reduced to the following form.4) The boundary conditions should also be expressed in terms of 4 as given in Fig.6) dx dy The two terms on the left hand side of Eq.177 where 4 = eCXT (A.5) Then Eq. which are functions of x and y only respec (x.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. c 2 X cx 2 = (A. A. tively. A.1.
3(B) where 0 is given by.= YO T 0 ii ii i h YO + co = 0 4 = 0.1.178 . Boundary conditions in terms of 4 for the problem in Fig. A. 5.0 0 = 0 x Fig. 4 = ec T .
A. A.179 The solutions to Eq.1 may be given in terms of the variables X and Y as follows. 13.11.11) dY dy=Y 0 = 0 (A. 13) In order to satisfy the boundary conditions given in Eq.9 and Eq.10) The boundary conditions in Fig.9) Y = C3sin8y + C4cosSy (A.15) ny 0 = n C2 _ n+c 2 e nX 1 n (a n 9 C) (A. A.8 can be easily obtained as follows. A.16) . X = Cleax + C2 eax (A.12 and Eq. The boundary con dition at x=0 will be considered later. A. A. the constants in Eq. dY dy y=O Y=O 0 (A.10 should satisfy the following relations.12) (d dx _cX) = 0 x=xO (A. C3 = 0 (A. Eq.14) (A. A.7 and Eq.
.) (A.y) = e cx (x.3 . the constants A. the final solution can be expressed as follows. 17) T(x. nw sin2 nO0 (n=1.19) .2. ax = A0e + C ax Z A [e n=l n a +c n n 2a x 0 n ea nx + ( a c )e e ]cOSSnY n (A.18) n n Since the boundary condition at x=0 should also be satisfied. Th 0 2. Therefore.1 are determined as follows.y) where nr (A. .0 2 A n Th h nr 1 n 1+( a c )e n 1 a +c 2a x .180 where 2 =c 2 + n (O) n 2 yO =c 2 + 8 2 .
4. Therefore.1Eq.1 and B. They are formulated for a natural convection problem where the gravity acts in the zdirection. The most important sweep should be given the first priority and most updated information.i. Since there are two steps involved in solving the momentum equation in each direction. the time step size in the ADI should be at least two or three times the Courant limit in order to have a gain in the overall computational efforts required for a problem. the computational effort for one time step is greater than that of the explicit scheme. When the tilde phase of the SMAC scheme and other scalar transport equations are solved by the ADI scheme. 181 APPENDIX B ADI SCHEME FOR MOMENTUM EQUATION The ADI scheme can replace the explicit scheme in the momentum equation in order to eliminate the Courant condition for the time step size. The ADI scheme is applied to a two dimeraional momentum equation in the following Eq. there are six sweeps in a three dimensional problem and four sweeps in a two dimensional problem. The sequence of the sweeps is arbitrary and depends on the problem under consideration. The time step size limitation in the ADI scheme comes from the semiimplicit treatment of the nonlinear convection term. Equations B. B. B.iillaY~i~~l_ ^YII^^~II ( .3 are the z and x direction sweeps of the x direction momentum .
2) n+l i+j * .2ii+ Ar 2 + Wilj+ .4 are the x and z direction The z direction sweeps of the zdirection momentum equation. ar .u n+l ui+2 2 n+l Ar2 +2u j n+l ui j .ui+ j At  n+l * u i+lj i+ .3) .2ui+. u n * u j i+ Ui+D Dt j + * n i+ j+ Ui+ j . Equations B.P z  n * RZij+ ij+ ij+ (B.2 and B.Wi+ Az j Ui+ j 1 ui+ j+l * .u Lr * n+1 .182 equation.Rxi+ jui+ jui+ j RX * n+l (B. 2 Az + Ui+ 1 P  n  i+j i+ jui+4j (B.1) * * n * *  ij+ At Wi+1 j+ ij+ + i+Ij+wij+ uij+Wilj+ 1 P n~ . sweep of the z direction momentum equation is treated as the last one in order to be given the most updated information.
B.2. B. w At n+l Wi.4 have given almost identical results with the explicit scheme for a time step size below the Courant limit. .1.3. and B.183 n+l i+ * wij+ * i n+l +lWij+ * n+l zW.4) Equations B.+ n+1 n+l Az 2wi + w _P z P0 zP g z RZ n+l ij + wij+ 3wij+ (B.
2. y and z direction mesh interfaces. are assumed to be zero. y and z directions. In the second part of the program the flow split is considered such that every outflow velocity is partitioned into four components in a three dimensional case in proportion to the velocities of the four neighboring surfaces in contact with the surface under consideration.3. The file. Then four crossflow diffusion constants at every interface are summed up to give a total crossflow diffusion constant .8. All the inflow velocities The component velocities are used to calculate the crossflow diffusion constants for eight subspaces in a three dimensional coordinate space.. 'difprg fortran'. 5. There are When there are inflow and outflow. the crossflow diffusion constants can be calculated in a simple way as in the first part of the program. Fig. is the program for testing Fig.8.3(A). 5. the calculational logic of the mesh interface implementation of the corrective scheme in section 4. are the computer programs for calculating the numerical values of the analytical solution in the problem geometries. two different cases treated separately. 'prog2 fortran' and 'prog3 fortran'. 'progl fortran'. or all inflows in x. When there are two outflows in any of the x.184 APPENDIX C COMPUTER PROGRAMS The files. a flow split occurs as shown in Fig.3(B) and Fig. 5.4. 4.
It is used to solve a one dimensional implicit finite difference equation in the ADI scheme. The total crossflow diffusion constant can be calculated in the same way for every mesh interface. The subroutine 'tdm' solves tridiagonal matrix problems by forward and backward sweeps. The ADI scheme for the momentum equation is given in Appendix B. the positive x direction.185 because one axis direction. has four neighboring subspaces. A portion of the VARR program is introduced to show how the ADI scheme is implemented in the energy and momentum equations. .
14 55 ATOT * THIODEXP(ALN*XO)/2.3F7.1 ) DX.NX.U.0 IHPI DO SUM 10 JJI.OALNOEXP(ALN.1415926536 C = U/(2.2) PI a 3.0 00 20 IPsi.NY  0.04ALP) CSO .ALP II FOIRMAT(F7.IALNIC)*DEXP(2.215. C*C XO " NXOX YO * PROOOO0 PROOOO20 PROOO30 PROO0040 PRO000OOOO P14000070 PRO000OGO P11000090 POOOOT100 PROOOO10 PIIOO0120 PRO00130 PROOO140 PROOO150 PRO00160 PROO000170 PRO00100 P1.NY) 100 FOnMAT(IOX.OZ) T(tO) DIMENSION REAO(10.3) SlOP END PRO00210 PRO00220 PROO0230 P1ZOOO240 P1I000250 PROOO260 PROOO270 PROOO200 PROOO290 P11000300 PROO0310 PRO00320 PRO00330 PROOO00340 PRO00350 PROOO360 PROOO370 PROOO300 PROOO390 PROOO400 .EO.NY.5)/NY ADD * ATOTTDCOS(YY) SUM T(d) v SUM + * ADO * DEXP(C*XO) 20 CONTINUE SUM 10 CONTINUE WRITE(G.0 44 YY I IIPI*(UJ0.100) (T(I).COF*DSIN(II*HPI)/(1.O*ALNXO)/(ALNC) COF * CCOFF/lI AAN .O) ALNX .N FILE: PROG FORrRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT REAL*0 (AH.0*TII/PI PI/2. AAN*2.YO) CCOrrF 2.2.26 II IF  IP  I GO TO 55 ALN = DSORT(CSO*II*II*YSO) (II.XO)/(ALNC) GO TO .4ALNX) ATOT .000190 PI00O200 NY*OX YSO v PI*PI/(YO.5F11.I=l.Tl.
0 X + XO/2. TL(10). GL(10).(ALN#C)*DEXP(2.5*YO . NY.DSIN(BEN+Y3) Y4C .0 SUM2 = 0.0 SUMaX SUM4 XN vI Y2 Y3 Y4 • 0.O) GO TO 55 ALNX .0 SUM3X 0.0 .0 .1.21 II = IP  I ALN = DSORT(CSO + IIII*YSQ) EXALO = DEXP(ALN*XOO. 1415926536 C .O*ALN*XO)/(ALNC) COF = CCOFF/II AAN . GH(10).5) EXAL = OEXP(ALNOX) XAL = DEXP(ALN'X) EXALN D EXP(ALN*XN) XALN = DEXP(ALNXN) IF(II.p A FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 MPI. ALP READ(10.5+YO 00 20 IPs1.3F7 .21)*DX = X 4 YO/2.NY2 0.0Z) DIMENSION TII(10).OTHIH/PI 11PI = PI/2.OALP) CSO = C'C XO * PROOOO O PRO00OO20 P R000030 PROOOO40 PROO00SO POOG0050 PROOOO000 PROOOOOO PROO0090 PIR000 100 P11000110 PRO00 120 PROOO 1.0  RTO DO =  OSORT(XXX) 10 JJ=1. 11 FORMAT(F7.X = 1.11) DX.O.0 X = (JJO.215.0 NY2 XXX NY/2.3.0 = YO/2.COF*DSIN(II*HIPI)/(I.0 SjU3 " = 0. U.U/(2.OCOS(OEN*Y4) Y4S DOSIN(3EN'Y4) ADO * AAN*(EXALO + ALNX*XALO)*YOO PROOO520 PR000530 P1OOO00540 PR000550 .0 = 0.FO.2) PI .DCOS(BEN*YI) YI Y22 • DCOS(BENY2) OCOS(BEN*Y3) Y3C Y3S .5)*DX = YC = (JJ.DCOS(BEN*YC) . TC(IO) TH14. IC1l REALi0 (AII. NX.5) XALO = DEXP(ALN*XO*O.2.0 SUMO SUMI r 0.XN = XN .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.+ALNX) BEN 11*PI/YO YOO .
NY2) = I. ALNX*XALN)*Y3S ADO3Y * ADD4X * AAN*((C+ALN)*EXALN + ALNX*(CALN)*XALN)*Y4C + ALNX*XALN)#Y4S AIDD4Y * EN*AAN*(EXALN GO TO 77 55 AOOI = TIIH*XAL*O.SFII.0 ADOvY * 0.II.I 1OO FORMAT(IOX.00) (GL(I).I WRITE(6.5) SUMI*DEXP(C*X) SUM2*DEXP(C'X) (SUM3X+SUM3Y)*DEXP(C*XN)/RTO (SUM4X+SUM4Y).0 ADO3Y * 0.NY2) WRITE(6.NY2) WRITE(G6.NY2) = t.(" % FILE: PROG2 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 ADOI w AAN*(EXAL+ALNX*XAL)*YII AD02 .EXALN + ALNX*(CALN)*XALN)*Y3C EN*AAN#(EXALN .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 .100) (GH(I).tOO) (TC(I).AAN*(EXALALNX*XAL)*Y22 ADO3X v AAN*((C+ALN).NY2) WRITE(.0 77 SUMI = SUMI + ADO SIIMO a SUMO + A000 SUM2 a SUM2 + A002 SUM3X = SUM3X + A00O3X SUM3Y w SUM3Y + ADD3Y SUM4X s SUM4X + AOO4X SUM4Y a SUM4Y + ADO4Y 20 CONTINUE C TC(dJ) THI(JJ) TL(J) Gi(JJ) = = GL(dJ) 10 CONTINUE C WI ITE(6.5 A)DO * TIHl'XALO*O.II.5 ADO3X 0.5 A002 * TMIIIXAL0.II.0 AOOX a 0.tOO) (TL(l). 1OO) (TH(I).3) STOP ENO SUMO*DEXP(C*XO0O.
O' XO/RT3)/NY2 XI=(dJ0.0 XC=Y0/(2.215.DCOS(UENOYC) YIl DCOS(OEN'YI) Y22 .S)*(XO/RT3YO/6.5)*(XOYO/1r3)/NY2 YC=(JO.0 SUM4X = 0. NX.EO.0) DO 10 00=1.0*RT3))/NY2 Y3=(JJO.0+(dJO.0 SUM3X = 0.SO5T(2.0*RTJ))/NY2 YIYO/2. U.O*NY2) .0 SUM3Y = 0. FILE: PROG3 FORIRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 IMPLICIT DIMENSION C 11 REALtO (AH.0 NY2 nRo P000O070 13100000 PROOOOOO PROOO0000 PRO00110 = NY/2.5)YO/(2 .0) PROO000120 PROO000130 PROO0140 PROOO150 PR000O160 RT3 = C SORT(3. G1(10). ( 10). NY.o) ALNX = (ALNwC)tDEXP(2.0*NY2) X3YO/(2.NX*DEXP(ALNXI))*YI I ADD2AAN*(DEXP(AI.0*RT3)+(J0.5)*YO/(2.0 .1415926536 C = U/(2.+ALNX) OEN = II'PI/YO YOO .(dJ0.2) PI = 3.OtTIIIM/PI HIPI = PI/2.0(JJ0.0 PR00O170 PR000100 PROO0190 PR000200 PlO00210 SUM2 = 0.0)/NY2 X2=(JJ0.4) Y4S = DSIN(6ENlY4) ADDO=AAN*(DEXP(ALN*XC)4ALNX*DEXP(ALN*XC))*YOO ADDIAAN9(DEXP(ALN*XI)+AI.5)*(XO/RT3YO/6.0*ALP) CSOQ C*C XO = DX'NX YO = DX*NY YSO = PI*PI/(YO*YO) CCOFF = 2. TL(IO). TIIl.2.5) * (YO/3. rORMAT(F7.0 SUMI = 0. ALP PROOOO10 GL( O).NY2 SUMO = 0.0*RT3#NY2) Y2=YO/2.0.O*RT3)+(JJO.0)/NY2 X4XOYO/(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).C.3F7.XO/RT3.5)*YO/(2.O*ALN*XO)/(ALNC) COF = CCOFF/II AAN = COF*OSIN(II'HPI)/(I.22 II IP I PR00360 PRO00370 PROOO3UO P(1000390 P000400 ALN = DSORT(CSO + II*laIYSO) GO 10 55 Ir(II.0 SUM4Y = 0.5)YO/(2.5)*(XOYO/(2.RT3)+(dJO. I 1) DX. TC(1tO) PROO0020 PROO0030 PROO0040 PnOOO050 PROOG000 READ( 10.S)#(XOYO/(2.N*X2)+ALNX')OEXP(ALN*X2))*Y22 AD003XwAAN((C+ALN)*DEXP(ALN9X3)4ALNX*(CALN)*DEXP(ALNtX3))Y3C PRO000410 PROO420 PR000430 1R'000440 P1100050 PR00O4GO PROO0470 PlOOO400 POO11000490 PROOOSOO 1100005 10 PROOO520 P0000520 P1OO0530 PROOO040 P1000550 .DCOS(iEN*tY2) YV3C = DCOS(BEN'Y3) Y3S = DSIN(BENtY3) YAC = OCOS(3EN*Y.*0RT3#NY2) Y4=YO/3.0.
100) (TC(I)..5*RT3)*DEXP(COX3) GHi(JJ)(SUM4X#O.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.NY2) WRITE(G.5 ADD2TtII#DEXP(ALN*X2)*O.11. 100) (GI(I).NY2) WRITE(6.1OO) (TL(I).5+SUM3Y*0.SUMI + P1I000620 PROO00630 PRO00640 SUMO w SUMO + SUM2 a SUM2 + SUM3X * SUM3X SUtM3Y = SUM3Y SUM4X SUM4Y * * ADO ADO A002 + ADD3X + ADO3Y SUM4X + ADD4X SUM4Y + AO04Y 20 CONTINUE TC(JJ)=SUMO*OEXP(C*XC) TtI(JJ)=SUMI.5 ADO3X PROOOSGO PROO00570 PR1000500 PROO000590 PROOOGOO P1100000.5 ADDH11H*DEXP(ALN*XI)*0.0 ADO4Y = 0.5*RT3+SUM4YtO.0 ADDOX a 0.I= . = 0.0 ADO3Y = 0.FILE: PROG3 FORTRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 AO3Y=BEN*AAN*(DEXP(ALN*X3)4ALNXODEXP(ALN*X3))*Y3S ADI)4XAAN*((CALN)*DEXP(ALN.X4) ALNX*(CALN)*OEXP(ALN*X4))*Y4C ADO4Y.I=.SF11.NY2) WRITE(6.5)*DEXP(C*X4) 10 CONTINUE WRITE(6.100) (TH(I).I=I.OEN*AAN(DEXP(ALN*X4)4ALNX*)DEXP(ALN*X4))*Y4S GO TO 77 55 ADDOTIIII*DEXP( ALN*XC)*O.100) (GL(I). PROOO6tIO .I1.3) STOP ENO .NY2) tOo rORMAT(IOX.0 77 SUM1 .NY2) WRITE(6.
5) DELR.0.OR.VELT(O).DIFII'IS2 rGR RfGTI .0 ) C DIFR .0.5*( VM V2 0.0) IF(VM.ISC.GT.O 151 4 0 ISS1 O 156 O DIFOOOIO DIFOOO20 DIFOOO30 OIF0040 011:O0050 DIFOOOGO DIF00070 DIFOOOTO DIF00110 RFGRI = 0.AOS(VM) + AOS(VP) AUS(WM) + AOS(WP) DIF00320 DIFOO00330 DIF0340 DIF00350 01F00370 D)1F0030 D roo390 DIF00400 DIFOO4 IO 1FOO00420 I) FO0430 DIFOO440 DIFOO450 DIFOO4GO )DIFOO470 D r004flO DIFOO4O DIFOO500 DIFO0510 DIOO00520 ODFOO530 DIFOO540 DIF00550 bMtt .t = 1 = i v 1 * .DIFTA(0) REA( 10O.I 01o00200 DIF00210 DIF00220 DIFOO230 DIFOO240 DIF00250 DIF00260 DIFO0270 DIF00200 OlFOO290 DIFOO300 DI F003 10 I'I .5+( UM U2 0.UU*DELR*( PPZ + PPT )/PSUM DIFZ * VV*DELZ*( PPX + PPT )/PSUM DIFT .O) IF(WP.5*( WP C IF( ISA.O.L1.O.OR.EO.VELZ(0).O.WM.DEL'TII.0) IF(UP.) ISA .WW4DELTII#( PPX + PPZ )/PSUM C RiFGII .0.0 IF(UM.UU/DFLR PPZ * VV/OELZ PPT .WW/DELTII PSUM .EO.O.GT.IFZA(0).5.DIFZ'IS3 * DIFZ*IS4 RFGr GO TO 00 AOS(UM) + ABS(UP) DI F00170 DIFOO IGO DIFOO00190 DIFOOlOO ISI IS2 IS3 IS4 155 ISG .GT.( WM W2 0.0 RFGCR 0.5*( VP WI * 0.IS .5*( UP VI 0.VP.DELZ.O 15S3 .DIFRA(0)1.0 n I FOO 130 OMTI m 0.WP 1SI * 0 152 .2.DIrn*ISI .O.2) UU * UI + U2 VV .0 OMT .LT.VM.0) IF(WM.2.0 RFGTI . bt ts .VI + V2 WW = WI 4+ W2 PPX .UP.LT.EQ.PPX $ PPZ + PPT GO TO 500 IF( PSUM.O.0) IF(VP.ISt + IS2 ISO IS3 + IS4 ISC ISS5 + ISG C UI = 0.0 RFOT 2 0.UM.EO.4 ~ FILE: O FI'IlG FOITIAN A VM/SP COINVERSATIONAL MONITOR SYSTEM PACE 001 DIMENSION VELX(0).
VELZ(I)/DELZ PPT . Z(2) vrI.NIE.*( PPX + PPZ )/PSUM GO TO 30 600 OIFRA(I) .0PFT) U I PFZ (I 1.FILE: DIFPRC'.0PFX)*PFZ DO 30 11.0PFX)*PrET V2*( 1.X(I)'DELR*( PPZ + PPT )0 'PSLJM DIFZA(1) m VELZ(I)*DELz#( rrX +.0PET) Vi*PFX*( 1.0 ) GO TO 600 DIFRA(I) .oPF7Z)'PFT UIs . 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.PFT) WI#PI'X#PFZ WI'*ix( .PrT )/PSUM DIFTA(I) .0.0.) U2 4PF Z PET 1J2'( 1.EQ.DELTH.VEU.0PFZ) DIFO00650 oDI roofGo oir00670 nir006s0 mrof0069 ot rooioo DIF00710o VEIX(S) VELX(6) VIELX(7) VELX(0) VEL7( I) V El.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 .PFT u i.V(LT(I).0.0PFX)o( t.0OPF Z) W2 *P FX' IF Z WI'.0 (Vl*V2).0PET) Vt .( .PFX*PFT V2* I'F X'PF T V2.0PFZ) W20( 1.0 oirZA(l) ..0PFT) U2*PFZ*( 1.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.VELX( I)/I)ELR PPZ .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$( t.0PFX)#(t.0PFZ)'$( .OPFX)O( I.IPFZ)s( 1.0.VELT(I)/OELTH PSUM a PPX + PVZ + PPT IE( PSUM.0PFT.o 'FX'( 1.09'FX)#(t1.0 PPX .0PFZ)*PFT U20( 1.0.NE. oi rooG to Dl F00G20 D1F00630 01F00640 Ut 'PFZ .0PFz) W2 &1EX'&( 1.0PFX)#PFT V2( 1.0PFX)'PFZ Wt'(I.0 Ir (W1#W2).( 1.0 oiETA(I) .0PFT) V t.0 0.0.O.( i.0 Pr T Ir (UI+U2).0PET) Vl*( 1.NC.
RFGT.RFGR.OMT I.7) 5 FORMAT(9F5.2) G FOIMAT( III .OMT GO TO 200 500 WRITE(G.2) 7 FORMAT( ItII .RrGT I.6F5.20X.I.511S ORRY) 200 STOP ENO DIFOi IfO DIFOI 120 DIFOI 130 oDIFO140 DIFO 150 DIFOI 160 DIFO 170 DIOt t0100oo oIro 190 DIFO1200 DIF01220 I I .2X. FILE: OIFPRG FOR TRAN A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 003 RFGT OMrt OMT = DIFZA(1) * DIFZA(4) a DIFTA(3) + OIFTA(4) DIFTA(1) + DIFTA(2) DIFZA(S) + DIFZA(0) 4 OIFA(D) 4 DIFTA(5) I DIFTA(6) DIFTA(7) D 999 WIllTE(GG) RFGRI .
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.I KN a (X(I() X(K) RE TURN END  X(K41)*COK))/O(K) VAR09Z350 VAflo9:160 VA1109370 VA11093001 VAI.XN) DIMENSION A(N) .NMI .P(400).X(N).o(N).SUarIouT INE TDM(A .D.0t4O01 X( I ) I~) 13( ) 0(i1) NM I *N I 00) 2 1I.C .C(N).O(N).NMI 1+1 )/0( I) P( 1+I) A( 141)#*C( I) P( N(I+ I) 0(111) 3 X(N) *X(N)/0(N) DO 3 1I.
O.F2(J.S( 4 .xx'..1) GO T0 49998 SOLUTiON FOR SIE.3.C .EO.G  V RIS150 VAR.) G0 TOrOOiC D'.5(V( K)*:(.K2NC JKZ2 . C C C C U AND v LDE VELOCITY EDU.LT.UG.KS 2.L.E230 .K) z 0.P2) GO TO 40002 (U(iK)*U(KX))) LUG = LS(O.K) • AVG..E0.KBP2 " (3.VGRD2/(.. D2).ItD.EO.]K .CR.PM.L!S( (]KZZ) ) . I O50 VI8090 K1 K12 2 KEP CONSTA'NT CLLCULATION C CRO5SFLD'.JSP2 DO O40000 DO 4000 K.)K2NC " (Ki)'NWPC .NPC o1 ]F(K.LT.:X GO TO 40000 "0OO D.C "hT..K t KXX a jK .P.LE0.S . iB10 .I.G.E0.I. DA'' G.0 OOC COI~'T2]'UE C VAR P182 vLR a210 VAR E220 VAR.K) * 0.TIONS SECT ION AD !ST ]F(ICLL].UG..0.FX(3.S(0. DNF.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.K) a LUGDX. 1) GO TO 400 IF(K. I vtR I5S0 rooOT 00G2 UC 'S(U(]K) ) VLRi 1210 vAR1E230 VAR E20 U(3KXX) ) LUGc .0 D'rF2( I .KZ2))) . 1CONV.FFX ( .E1O 0005 FP" * IF(.2D LUGRDX/(AUGOX +~DG'RD)' ..DFFUSI'O =.
054 (K. KS I KLS a3)K I KP a IK  NP 2 I N1PC VAR 185501 YLR 15560 VAR 1C570 VARias590 VAR 86201 VARi1Es30.E. UD( 3K) UL UR UT Uo PK) UD( 3KP) U0( 3KMS) DCR DCL D!FFCD(IKL) D) F CO( KL. ( '(K ULH C CON'.PKKNUO(IPK) ).1 SRKSrI a * N BUOJY a LP0E7.1.5 PC) ').0.P( TJ"'.ET/TADD3 VAR IS420 VLO IS440 3F~lDw.E003 DO E0100O ) K GO TO50205 222....0.I DIFFCD(3KL2) D!~FFCD(IKL3) *0(!.I ) I KIPC # (KI)'tN'PCL VAR16D VAR 11510 VARi18520 !PK a 31. .5 UODUMKS) 16D(1PK) * .2)GO a TO 29990VALR iS.I) K2WC' + II)K NC *K2NC "00Z0100 K=2. a uJD(2K) )0.5 )r0. VLRIE6t0 V4RI15650 VAR 15690 ULWM'(uLLIC) ( IR'(L'C4UV) VARIE *7 DO.C RD20 3KS0 S3G'~ to VAR 8340: VARIS5360' RDZDRD2 VAR i !:0.'O('.S(UftH)(UCUQ)  VL21S660.KSPI I KL I lK z lK 1A 0.R2CLI.]BR ( '. 3F(3~.
=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 )*.'N(tX) + 06Md~7A 1 Et a 6Nr NZ At 2 SW SWA r r )..uQ + + t (Nt)0ftl x = XIt)f ( oo.(iim1 (.8~ VA 0899tMVA (oaQN 3NC647.(.7A 0060tMIA 0L6Sgv7A 0968LVA I 3a CA~ SimI) 0L96 t'7A O9S63tM7A j0 0 VA d061Nt) AoflsaNx0As + Od.+fl)IdM Y(Qd.(onin) M0'a(On.(?nOn) i13( fl2f) unc.t) 2 td9N'Zt oo0to oa 3flN tLNOD s0?_O3 M 6 dVA + NAcns + ZntO * (Nt)d(dXt)d ).) .o.o2m.)10 6666a 01 0 MfN i NGO 07 6 "VA ZNOO NCO ZOM( + (Xt)0M £ ) O~t6tVA (~ ~ .. (W (4I) 09O6t~u7A CA ~g0 *0 H IL~ Iun S:)o ( (dN I)an +(>(!M)Onl 'S0 06SLIA OEOG bVA 0L6O t'7A 0CS8~ 100 VA )O (dm)OoA1 0669'.Cw'))'A.( (nf1l)(Himlss7 (In3n)(uLM)S9V ) ZNQ.vt Z1t0 7 Z(J~Si0(o(nsn)(a)a (1n.OSZ6 M7A OdI'N Al dN O7I~t 17A OdM ( A) ONZ Nt + 1IOZN4(Vt) * + 3NZ)(11) + t bi > >4 E A c TA .onH.% At t x NdE ONCA + lA b + A t + OZ~N(Lt) OdAN 00gla aVA 06!.
ROT * 0.SIGIArRDZD3DCT 01(J) a U(IK)ROT 81000 CONTINUE C .IMKS IKt.1 ]K £ + (1])K2NC + (Ki)NWPC ) VAR 19260 VAR 19270 VAR19 280 vLRlg2O VAR 19290 VAR19300.SIGMARDXD*DCL ACBS(URT) .5RDZ*( VRT * ASS(wRT) .'RDXDDCR D01(J) .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) ) .ABS(VRT) ) .5 AI(J) a 0.( DCL * DCR ) C1(J) z 0.RDX.]K .VRE + ABS(WRB) 1 * SIGMA*RDZD.C1 .ULT * ABS(ULT) 81(J) .KBR) C DO 82100 Ka2..1SP J a (K2)]IeR * 1 .SIGMhARDZDODCB E1(J) = RDT * 0.KBR DO 62100 1=2.'5 WRS (W(1KMS) * W(IPKNWPC))O.C1.5( U()]KS) * U(IMKS+NWPC) ) Ai(J) = 0.5.KER D00 82000 12.KSPI ]K = 1 + (11)K2NC + (KI)NWPC 1X a (12).IKBR) CALL TDM(A 1.8e1.AeS(WRB)) .DI.X.5( U(IK) + U(]KP) ) ULT C a 0.IBR DO 81100 K2. C 00 l8100 I=2. = DIFFCD(IKL+3) WRT 3 (W(3K) + W(iPK))O.KBR + K 81100 U(IK) a X(IX) C DO 82000 K=2.5RD02( WRe .0.5RDX( URT 1 + SIGMARDXD.$ ]K .W(1K)RDT 82000 CONTINUE C CALL TDM(A .DIFFCO(IKL) DCL a D0FFCD(IKL+2) URT x 0.]K + K2NC IKP a 1MKS 1K + NWPC ]K K K2NC NWPC 1KIAS = C DCR .SIGt.ABS(ULT) ) .( DCT + DCS ) C1(J) a 0.5*RD2( WRT .NWPC DCT 3 DIFFCO(IKL+1) DC.5*RDX( ULT .X. VAR19310 VAR19320 VAR19330 VLR193AO VAR19350 VAR19360 VAR1S370 VARie380 IKL a C 1 + (1)'K2NCL + (K1)NWPCL IPK .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 .K2NC .
3)3QO. z (r 01)~S~7~).LG!1CE)PA HIM.Htin 0~Q0~*/'.)ssv .(HL')S.fl.L 0.3 00 000O~d7A "L661 aVA (AI  (Ad 1) d !OSGOZi VA OG61 !VA 03661 SVA b33aXac~tjS ( 100 (Hll)SV H~ln   C(Hzn)sev  Ya   ~ A +Laa Hfl )xCi~s~o = x (r)La ()9 t ()1 2 (H.40 C0~7A O CO0d7A NAoC1s (A(I)a .ZC(3aE1~S (d~t)d  00oo' ((HIiA)SS7 I )'ZQZ .mNtAz + + ONA(lt.Q ( (S'feXtiM 094=J7A OSZO0t7A (>I).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( . )'ZQdS50 )Zc!S*  a ' a  (HS.0xCd..~) Al NP 1=I000ca 00 .v%4!)iS +.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 ).'%)SS'7 + HSA .(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.~tSC(Hifl)ss b3(3 ).3nwNtJ.' ).v4)tS ~m * io1. ).
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 .2.5(TS(IK)+TS(IPK)) .25*RRC CYRX = CYLO0.B.0P.EO. VAR20420.S]EO(]K) TSRA TSLA . IBR(K2) + I ' (1i)K2NC IK = i + (K1)NWPC IKL a 1 + (11)K2NCL + (K1)NWPCL RC a FLOAT(I1)bDX .K2NC ]K IK 7]KS IKMS = IK .1 (12)(KER1) 84100 W(JK) a X(lX) C 89999 CONTINUE IF(ICONV.Ci.O.DIFFCO(1KL*i) DCL a D!FFCO(1KL+2) DCB * D]FFCO(JKL*3) ]PK )KP a a + K2NC * NWPC = 1K .12 J .S) SIECO .DASS(UC) AUL .I) GO TO 55000 DCR .1CALI.CF(UK) IF (CFC.U(JMKS) vW = W(IKMS) AUC .HDX RRC a I.DIFFCC(IKL) DCT .NWPC CFR = CF(1PK) CFL .1) GO TO 2040 ISEO = ISEO + 1 IF(ISEO.D1.NSEO) 33333 CONTINUE 15E0 a 0 DO 60000 K=Ki.EO.1SP1 00 84100 K2.CALL TDM(A .NE.LE.0/RC OR = CYL0.0.SIE(IPK) S*EL .KSR IK IX a I + (I1)*K2NC + (Ki)*NWPC + K .DLeS(UL) ALC = DAES(WC) 1wS * DASS(WE) = SIE(1K) S!EC SIER .X.IKBR) DO 84100 l=2.SIE(IKP) S]EB * SIE(]KW.5(TS(]K)+TS(lMKS)) VAR203G0 VAR20370 VAR20380 VAR20390 VAR20400 VAR20410 .SIE(IMKS) SIET .CF(]MKS) UC c U(IK) WC = W(!K) UL .5RRC*RDX CFC .K2 DO 60000 13a1.
L.(r'."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. (nN*31.OlI)03(r v+n .OSrlth7A OV7 t!VA t. N *:dM o zi Z)q4s)4 coo 0.1! Al z(IJ)L3 (NOfl'Ol V(ISJX017) Ot)t0dVA 0001ZlA N7 dlxiol ((L1)x~JNG4(IV~~sNQ)ts0 .I vo 0009 0:9>4 I )c t a.C*~ t )4* . 10 + ( 0S669 01 00 (V03VIJfl S0. t :) a .Q ~ 000 fl co(3 S6 .NOD (L03e..At~VA OSZI~7A clt'L fa ZAVtAZ'4 0=0i9 00 0=O9 00 I .± 4:7) 21J3 1u L7S1L'q'l OS 60ZMVA (Slx () I s I)SO I ((diE)l+ m1)l)'c = vzs I VIL .+c a 21J10 to.x)j! 2(AI1)3rS oo0t 9 06CI ~1VA' 09ZI dVA 0LI.(N t0 *090~bVA 0O0I bVA 0O'0t MVA OSOI aVA 0OP0tZSVA (Q QtlY. 06C Zd7IA ZltrEzf 00 OOOL CO 0SCtLb7A NOI0t1D3JCZ NE d33MS CNZ 0L~tZZ7A LCECE OZ :tZ7A 1 05 3 3lNTI.~t J '~1f ((t0AX'1ZJANO ) C dt0* ) * N ) S 4~ I . C0 r~t (r)LC 00~ttIdVA 060k ~wVA 0OLt dVA 00009 01 13t5. t7) .
X.(WC*AwCWS*iWB).1) GO TO 65500 D1(J) * 01(J) .5(DNFF2(3.0.0 E2(J) .1.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.E0.i) GO TO 70000 IF(CFT.I) )XMX GAMTTSTA VAR21620 VAR21630 VAR21640 VAR21650 VAR216GO0 VAR21670 VAR21680 VAR21690 VAR2I700 VAR217 tO VAR21720 VA.5(WCWC)...CO(IK)0.C2(J)SIE(IKP) C2(J) = 0.K.B2.NE.IKBR) C DO 71000 ]=m1.X GtA. 1 * 0.5(rS(lK)+TS(lKMS)) C GAT DIFFT DIFFB C C * RDTSIE(]K) .5"(ws*Aw )'RD2 .DIFFTDCTRDZD C2(J) a 0.5.(A2. .OlFFTDCTRDZD IF(CFB EO.0 GO TO 70000 65500 C2(j) E82(J) .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 ..RDZ C2(J) = 0.5 0.NU) GAMT * RPRAN .5(DNFFZ(I.LE.K)*DNFF2( .01.1) GO TO e9950 Dl(J) = 01() .C2.12 DO 71000 KKI.K2 Ix = KS.0.(12) + K .11.0.0 DI(J) D(J) + DIFFTDCT*RDZ GO 70 70000 65000 01( ) = SIE(]K) L2(j) a 0.IF C DCR OCT DCL DC8 C ( CFC.k)+DNFFZ(1.NNOPT.ND.5(TS(]K)+TS(3KP)) TSeL .K2NC IKWS VAR21460 VAR21470 VAR21480 VA021490 VAR21500 VAR215 O VAR21520 VAR2rS30 VLR21540 VAR21550 VAR21560 = ]K .I DI(J) a TGA.K+1))X.DIFFEDCRD2D RDT * (DIFFTDCT+D)FFS*DC6)*A02D .EO.RDZ .1) GO TO 65000 .TTSPLL .0 69950 IF(CFT.DIFFCO(IKL) = D]IFFCO(IKL+1) = DIFFCO(]KL+2) .M * * IF(TS(]K).A2(J)SIE(3KMS) A2(J) • 0.0 70000 CONTINUE C CALL TDM.EO.0 C2(J) = 0.NWPC C CFT = CF(IKP) CFB .DIFFCO(IKL+3) 1PK a 1K + K2NC IKP a !K + NWPC IMKS a 1K .
IK r 1 + (II)*K2NC + (K1).K2 1K 1 + (]i)'K2NC + (KI)*NWPC 33338 SIE(K) = 0.K) ) KIRK a 0 33329 CONTINUE C C NOTE.K2 IK v 1 + (1I1)K2NC + (K1)*NWPC STOR(I.SIEO(IK) KIRK a i GO TO 33334 33337 CONTINUE DO 33338 1=1112 DO 33338 K=KI.12 DO 33335 K=KI.NwPC 71000 SIE(]K) = X(IX) C IF(KIRK.1) GO TO 33333 GO TO 33339 IF(ISEQ.S)=W(KK+) TO(KMS)=TO(KK+1) VAR22470 VAR22480 VAR22490 VLR22500 VAR22510 VAR 22520 VAR22530 VAR225ZO VAR22550 .5'( SIE(IK) + STOR(I. 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. 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.E0.09 KzK1. TRANSFERS VELOCITIES TO STORAGE ARRAY( 2040 KI1 K2xKBP2 LWPC=I .NE.O) 00 33335 I11i.11.K2 LWPC=LWPC+NWPC AT TIMEN ) .NWPC DO 2.K) = SIE(IK) 333335 SIE(IK) .
APPENDIX D TYPICAL INPUT FILES The file. The file.3(B).7. 5. is the input to test the stability of the ADI scheme for momentum equation. are for comparing Huh's And De Vahl Davis and Mallinson's correction formulas in the problem geometry. 5. Fig. Fig. is for calculating the natural convection flow field in a Cartesian closed compartment. The files. 'input mom'. The output results are in Fig. 5. 'inputd 3' and 'inputd 7'. 4. are for calculating the numerical solution by donor cell scheme in the problem geometry.8.2. . results are in Fig.1 and Fig. 5.6 and Fig. 5. 'inputn dat'. The output The files. 'inputd HUH' and 'inputd DM'.
0 0.013 0.9 0.0 10.0 10.0 2 5 10.0 4 0.01 0.0 2 l.0 1.44 0.0 60.0 0.0 0.4 5 1 1.0 0.0 0.0 0.75 10.7 0.0 0.0 0.0 0.5 6.0 0.0 0.0 2.0 1.0 0.0 1.0 0.5 0.0 1.4 1.4 0.0 0.2 10.0 0.0 0.0 1.0 1.0 60.0 1.0 1.01 0 1.0 0.0 .0 0 0.0 0.0 0.0 0.0 4.0 I 3 I 16G 1.0 60.0 0 0 1.0 0.0 0 2 1 3 O0 2 2 O O ADI 10.0 0.t20 1.0 0.0 1.09375 0.0 10.0 0.4 0.0 0.0 1.0 0.0 1.045 0.rTh FILE: INPUT MOM VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 4 4 O STABILITY TEST FOR MOMENTUM 2.0 1.0 0.120 1.0 0.0 0.
0 1.0 68.0 1.01 1.44 0.0 677.0 32.+20 60.0 60.20 2 1.0 0.0 0.0 0.0 NATURAL CONVECTION IN A CARTESIAN CLOSED COMPARTMENT 1.2 1.20 1.0 0.0 1.0 1.0 4.0 2.4 0 0 0 0 O 0 0 6 2 0.0 10.0 1.90 0.0 1.0 0.7 1.20 0.0 0 0 2 1.67 1 1I 2.20 1. 20 1.5 0.0 1.01 0.5 1.5 5 1.045 0.21795 1.0 .0 0.20 030 1.0 I 43 1.751.0 1.20 II I 1.FILE: INPUTN DAT A VM/SP CONVERSATIONAL MONITOR SYSTEM 'PAGE 001 5 10 0 1 0.0 0.0 0.0 8 2 I 3 1 3 0 3 0 1 1.0 0.0 6 6 50.0 0.0 1.0 0.0 0 1.
0 1.457 1.CASE 3 2.497 0.0 0.000 10 1.026 5 1.207 0.0 1.0 0.0 0.47.000024 60.061 1 0.0 0.0 9.5 1.5 1.0 0.01 0 0.999 1.0 0.0 0.0 0.0 9.911 3 4 1.0 GO.0 0.0 1.+20 I 0. 020 2.0 7.75 0.207 0.0 5.0 7.0 0.0 0 0 NUMERICAL blFFUSION .0 0.0 9.70711 1.0 1 .05.0 0.000 0 0 0 O .0 10.045 1.204 0.063 ( ).5 0.0 0.70711 0.0 2 1t 1 113 5.249 3 1.0 7.014 0.0 7.000 0.0 1.997 7 1.0 1.0 0.409 1.013 2.0 0.0 0.0 0.940 1.3 1 I 1.0 1.0 0.014 ).0 9 10.991 6 1.0 0.0 0.0 4 I 36 1.0 0.07107 7.0 2 1.000 0.001 0 1.0 1.000 10 1.0 0.0 1.07107 0.0 1.0 0.0 1.0 1.0 5.0 9.457 1.0 0.0 2 4 2 4 1.0 12 11 2 121 t 5.0 7.0 0.0 0.000 1.9 0.0 5.0 0.497 2.0 60.511 1 1.07107 111 3 1 1 2 5.0 0.07107 1.0 0.0 0.0 1.002 0.204 0.000 9 1.02 1.974 5 1.0 9.0 0.0 9.002 0.0 8 10.0 1.0 1.0 1.0 0.948 1.0 0.003 7 1.0 1.751 2 1.0 1.0 1.01 0.07107 12 12 Ill 2 5.07107 0.000 0 0 O O I.09375 0.FILE: INPUTD 3 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 10 10 0 0.0 .II I 11 2 2 1.07107 0.0 0.44 4.000024 0.009 4 1.0 0.0 0.O.O 2 1.0 0.0 1.OGI 2.063 0.0 0 2 5 3 0 7 6 1.009 G 1.0 0.
0 1.0 0.0 1.01 2.44 4.0 0.0 10.75 2.0 0.0 1.014 0.633 1.0 0.0 1.07 07 0.0 0.911 9.001 0.0 7.0 9.333327 1.0 0.0 0.0 2 7 8 5.0 0.0 0.0 0.0 0.002 5 1.0 1.0 60.990 5 1.5 1.0 2 9.009 1.0 0.00.07107 711 8 8 2 5.0 1.0 1 9.0 1.0 2 2 0.0 0.0 0.0 0.0 10.000 6 1.07107 1 1 713 2 5.0 7.0 1.0 .0 5.0 0.0 4 2 4 1.000 0 O 0 0 0 7 1 2 7 2 5.003 0.0 1.0 4 1.07107 0.0 0.0 1.9 0.0 0.0 0.0 0.0 4 I 36 1.0 0.420 1.+20 2.0 I 1.0 3 0.0 0.457 2.0 0.0 3.0 7.001 0.0 1.0 0.01 0 0.0 1.0 7.17051 0.09375 .000 6 1.00.071107 0.907 3 1.CASE 7 0.01 0.0 1.5 1.014 0.FILE: INPUTO 7 A VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 001 6 0.0 0.374 1.633 1.07107 I 113 7 2 1.0 1.0 1.045 1.333327 60.0 0.47.0 60.0 0.0 1.4 8 2 1 3 0 0.3 1 1 1.143 0.003 17051 1.0 0.07107 7.0 0.457 2.167 1.626 1.0 5.000 0 0 0 1.0 0 0 2 2 1 0.0 NUMER ICAL DIFFUSION .0.4 1.0 0.0 2 0.0 5.013 2.0 7.0 0.013 4 1.0 1.143 0.
0 GLO10 ZU10 SL9*0 109.0 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. 6IP'0 O1'1 991'0 LOCO0 9LV'O 060*0 Cc£00 910'0 1 0 C Iz 0 0 0*1 OL+*t Ylz OZ4*1 zlo' IfIAH .09 0.0 01l 0'00' C0'I C.NOIS(fJJIG IV:)IUIW(IN too 39vd IVNOIIVSU3ANIX) dS/W~A v VIIi III)II OIMINI O~N :311A II .SAS 3~~~)Vd kOIINOW 0 0'0 0*0 0.0 09G"0 9Cs£0 661''0 6sv1'.09 W '09 0.0 0.0 CGO.1 1' 0*0 0 *1 D99C00 0'1 0IO' 0*0 0*0 0*0 0'0 0 *I0*0 9c 1 t.. 0 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 LG0 .0 9?LL '0 G'.L Z01 'L L'09C91C P09c91''? 1'Q9C9p 0'19C91' 1 iLO'O 1'LO*0 GL00O LLO'0 £00. 1'0 0.z 10'o 0 1 0'l GL'O 1O1 '0 I 0*0 01l 0'l L910* 199*0 099.*I riI 00 0*0 0*0 0i'00 0 10*0 6 v 0*0 0'0 top' . *0 PI 9.1bl b 690*6 £00*0 LC9 0 90c.u *0 Ul' I 6P I 0011* I 0.0 0 *I O'l 0' *1II1 01l 0*1 0 1 0'L 01 O'1 0.
20 I13 1.20 2 1.20 9 1 1.20 1011 1.20 2 1.0 0.0 5.0 0.0 5.0 0 2 1.20 1 1.20 5.0 0.0 2 25 5.0 .20 913 1.0 26 26 5.FILE: INPUTO IIU4U VM/SP CONVERSATIONAL MONITOR SYSTEM PAGE 002 2 0.0 8.6603 0.20 10 1.6603 0.0 I 1 5.0 2 25 25 5.20 911 1.6603 0.
0 0.026 0.0 1.072 ' II 12 13 14 15 16 17 10 19 20 21 22 23 24 0 0 0 I '.567 9.099 0.01 0 0.33G 6 1.0 0.02.3 1.0 1.2406 1.013 2.0 0.003660 60.0 1.0036G0 0.75 1.496 0.365 9.0 1.0 1.0 0.156 0.045 1.223 3 1.01 0.030 0.0 0.705 0.0 0.0 0.420 0.000 0.246 0..0 0.0 1.0 1.0 0.5 1.rILE: INPUIO OM A VM/SP CONIVEIlSATIONAL MONI(fro SYSTEM PAGE 001 24 f( 0 I 0.0267 0.463604 2.302 0.003 0.002 2.526 3 4 5 G 2.30i9 1 0.44 4.076 0.093 0.4GI6 0.5 1.0 1.030 0.260 0.021 0.0 0.OM 0.0 1.009 0.120 2.0 1.0 0.0 1.003 0.566 0.0 1.0 0.460 4 1.0 1.002 0.0 1.02.174 1.920 0.0 1.0 0.949 0.0 0.0 0.0 I 1.0 0.20 1.0 5.0 1.OGI 0.463G04 2.0 0.0 0.115 0.0 0.0 0.0 0.034 0.0 5 1.767 0. 1.242 7 1.0 1.197 0.051 0.133 10. 4G1IG04 7.0 3 10 1.0 1.0 0 2 1 3 0 7 6 I 0.700 .0 1.0 1.951 0.029 0.969 0.4 0.016 0..522 0.4 4 1 36 1.463604 2.0 1.360 0.0 0.0 1.144 1.0 0.0 1.0 0.0 1.932 1.0 0.0 1.02 2.463704 9.0411 0.044 9.403604 2.5 1.636 0.0 0.0 0 0 NUMEtICAL IrFFUSION .0 2 4 2 4 1.0.0 0.GOG 2 1.9 0.09375 0.0 60.005 0.0 0.0 1.0 1.011" 0.007 0.0 0.0 0.0 0.903 0.002 0.0 1.003 0.0 2 2 2 0.0 0.0 60..0 5.96G 0.
0 I 5.0 2 5.20 113 1.6603 0.0 .0 5.0 0.0 0.20 913 1.0 0 29 25 I 25 2 1.20 I 2 1.: INP. GG03 0.20 1011 1.0 0.20 9 1 1.0 5.SILr. & .20 5. 6603 0.20 1.0 I*  .0 2 5.20 911 I.0 26 2 1.0 26 5.10 OM A VM/SP CONVERSATIONAL MONITOnr SYSTEM PAGE 002 2 0.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.