CFX-4
Introductory Training (Dec 1999)
8 -1
CFX-4.3 V4.3.1H
Lecture 8
CFX-4 &
Output File
CFX-4
Introductory Training (Dec 1999)
8 -2
CFX-4.3 V4.3.1H
CFX-4 Basic Concepts
? General form of equations solved in CFX-4
? Integrating the equation over the control
volume
transient convection diffusion source
SUt ?????-??? )()( ????? ? ? ?
? ? ?
? ?? ?t dV U n dA n dA SdV? ???? ? - ? ?
? ? ???,
CFX-4
Introductory Training (Dec 1999)
8 -3
CFX-4.3 V4.3.1H
CFX-4 Basic Concepts
? All terms are discretized in time and space,
linear equations are obtained
Discretization
> Transient term,backward difference,Crank-Nicolson
> Convection term,Hybrid,HUW,QUICK,CCCT,etc.
> Diffusion term,2nd order central differencing.
The linear equation is spatially coupled,Direct
matrix inversion is too expensive,so iterative
linear equation solvers are used
> ICCG,AMG,Stone,line relaxation,etc.
A A Sp p nb
nb nb u
? ?- ?? p N C E L L? 1 2,,,?
CFX-4
Introductory Training (Dec 1999)
8 -4
CFX-4.3 V4.3.1H
CFX-4 Basic Concepts
? CFX-4 uses a segregated nonlinear solver
(SIMPLEC),Each nonlinear equation is solved
independently of others,
initial guess
final solution
U momentum equation
V momentum equation
W momentum equation
pressure (mass) equation
…,..
sweeps
(inner
iterations)nonlinearouter
iterations
convergence
CFX-4
Introductory Training (Dec 1999)
8 -5
CFX-4.3 V4.3.1H
About Convergence
? Convergence:
the solution has reached an unchanging state.
in CFX command file,the convergence criteria is
set with the mass source tolerance.
the mass source tolerance is set as a percentage
of total mass flow rate.
the solver defines convergence when the mass
source residual is less than the mass source
tolerance.
the mass source residual is NOT dimensionless,it
is the sum of the absolute values of the net mass
flow rate into or out of every cell,
| |ri
i
ncell
?
?
1
CFX-4
Introductory Training (Dec 1999)
8 -6
CFX-4.3 V4.3.1H
About Convergence
a mass source tolerance of 1.0E-3 x total mass
flow rate corresponds to < 0.1% mass imbalance
in the system.
for convergence,residuals for enthalpy and other
scalars should also decrease by a reasonable
amount,and the monitoring point values not
change.
the flow field CANNOT be expected to be
physically meaningful if convergence has not
been achieved.
CFX-4
Introductory Training (Dec 1999)
8 -7
CFX-4.3 V4.3.1H
About Convergence
? Good convergence ? Poor convergence
CFX-4
Introductory Training (Dec 1999)
8 -8
CFX-4.3 V4.3.1H
About Convergence
For transient problems,a converged solution
should be obtained for each time step.
CFX-4
Introductory Training (Dec 1999)
8 -9
CFX-4.3 V4.3.1H
Output File
? The CFX output file,m##.fo,is an ascii file
which provides numerical data on the
calculation.
? The user can browse the output file for
important information.
CFX-4
Introductory Training (Dec 1999)
8 -10
CFX-4.3 V4.3.1H
Output File
? The output file by default contains:
1),Front-end printing
physical models & numerics
model topology
real boundary conditions
2),residual and monitoring point values
3),flow field variables at the end of the calculation / time step
4),wall information at the end of the calculation / time step
5),flow boundary flux information
6),summary of convergence,workspace and CPU time used
*
*
CFX-4
Introductory Training (Dec 1999)
8 -11
CFX-4.3 V4.3.1H
Output File - Example
? Residuals and monitoring point values:
TOTAL FLOW INTO REGION THROUGH INFLOW/OUTFLOW BOUNDARIES:
INLETS = 3.046E-06
OUTLETS = -3.046E-06
MONITORING POINT AT (1.0700E-01,4.7600E-02,2.7500E-02)
ITER I---------ABSOLUTE RESIDUAL SOURCE SUMS ---------I I--------FIELD VALUES AT MONITORING POINT--------I
NO,UMOM VMOM WMOM MASS ENTH U VEL,V VEL,W VEL,PRESS,ENTH,
1 0.000E+00 5.500E-12 3.040E-09 6.091E-06 6.300E+00 1.165E-03 6.525E-04 7.970E-07-1.377E-04 6.583E+03
2 4.829E-07 3.348E-07 3.573E-07 4.865E-06 4.554E-01 1.327E-03 7.478E-04 1.200E-07-6.969E-05 1.355E+03
3 1.727E-07 1.257E-07 1.327E-07 2.140E-06 9.279E-02 1.432E-03 8.013E-04-5.941E-07-4.079E-05 2.242E+02
4 8.008E-08 6.050E-08 6.046E-08 1.229E-06 2.313E-02 1.510E-03 8.401E-04-1.126E-06-2.921E-05 1.862E+02
5 4.828E-08 3.675E-08 3.646E-08 6.667E-07 1.326E-02 1.567E-03 8.683E-04-1.513E-06-2.415E-05 1.784E+02
.,,,,,,
.,,,,,,
75 4.830E-13 2.388E-13 6.643E-13 3.759E-11 1.482E-06 1.732E-03 9.410E-04-3.916E-06-1.778E-05 1.650E+02
76 4.177E-13 2.107E-13 5.213E-13 3.288E-11 1.461E-06 1.732E-03 9.410E-04-3.916E-06-1.778E-05 1.650E+02
77 4.159E-13 2.073E-13 5.951E-13 3.373E-11 1.426E-06 1.732E-03 9.410E-04-3.917E-06-1.778E-05 1.650E+02
78 4.470E-13 2.217E-13 3.778E-13 2.846E-11 1.443E-06 1.732E-03 9.410E-04-3.917E-06-1.778E-05 1.650E+02
AVERAGE REDUCTION FACTOR ACHIEVED BY LINEAR SOLVERS
U MOMENTUM RESIDUALS 6.0E-02
V MOMENTUM RESIDUALS 8.4E-02
W MOMENTUM RESIDUALS 5.2E-02
MASS SOURCE RESIDUALS 2.7E-01
ENTHALPY RESIDUALS 6.2E-01
CFX-4
Introductory Training (Dec 1999)
8 -12
CFX-4.3 V4.3.1H
Output File - Example
? flow boundary flux information:
* * * * * * * * * * HEAT FLUXES ACROSS WALLS * * * * * * * * * *
TOTAL CONVECTIVE HEAT FLUX ACROSS ALL WALLS = -3.501E-02 W
* * * * * * * * * * MASS AND ENTHALPY FLOW THROUGH INLETS * * * * * * * * * *
PATCH NAME,INLET1 BLOCK NAME,BLOCK -NUMBER-2
HIGH K AT K = 10,I = 1 TO 10,J = 1 TO 10 PATCH GROUP NUMBER = 1
MASS FLOW = 1.218E-06 KG/S,ENTHALPY FLOW = 2.081E-02 W
PATCH NAME,INLET2 BLOCK NAME,BLOCK -NUMBER-3
HIGH K AT K = 10,I = 1 TO 10,J = 1 TO 10 PATCH GROUP NUMBER = 2
MASS FLOW = 1.827E-06 KG/S,ENTHALPY FLOW = 1.285E-02 W
TOTAL MASS FLOW THROUGH INLETS WITH PATCH GROUP NUMBER 1 = 1.218E-06 KG/S
TOTAL MASS FLOW THROUGH INLETS WITH PATCH GROUP NUMBER 2 = 1.827E-06 KG/S
TOTAL ENTHALPY FLOW THROUGH INLETS WITH PATCH GROUP NUMBER 1 = 2.081E-02 W
TOTAL ENTHALPY FLOW THROUGH INLETS WITH PATCH GROUP NUMBER 2 = 1.285E-02 W
CFX-4
Introductory Training (Dec 1999)
8 -13
CFX-4.3 V4.3.1H
Output File - Example
* * * * * * * * * * MASS AND ENTHALPY FLOW THROUGH MASS FLOW BOUNDARIES * * * * * * * * * *
PATCH NAME,OUTLET BLOCK NAME,BLOCK-NUMBER-4
HIGH I AT I = 15,J = 1 TO 10,K = 1 TO 10 PATCH GROUP NUMBER = 1
MASS FLOW = -3.046E-06 KG/S,ENTHALPY FLOW = -8.762E-07 W
TOTAL MASS FLOW THROUGH MASS FLOW BOUNDARIES = -3.046E-06 KG/S
TOTAL ENTHALPY FLOW THROUGH MASS FLOW BOUNDARIES = -8.762E-07 W
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
RATIO OF RESIDUALS FROM SECOND AND LAST ITERATIONS
U MOMENTUM RESIDUALS 1.1E+06
V MOMENTUM RESIDUALS 1.5E+06
W MOMENTUM RESIDUALS 9.5E+05
MASS SOURCE RESIDUALS 1.7E+05
ENTHALPY RESIDUALS 3.2E+05
TOTAL REAL WORKSPACE USED = 839097
TOTAL INTEGER WORKSPACE USED = 325039
TOTAL CHARACTER WORKSPACE USED = 537
TOTAL CPU TIME = 91.2 SECONDS
CFX-4
Introductory Training (Dec 1999)
8 -14
CFX-4.3 V4.3.1H
Output File
? The contents of the output file can be controlled
by the command file:
.,,
>>OUTPUT OPTIONS
>>FRONTEND PRINTING
keywords /* frontend printing control */
>>PRINT OPTIONS
>>WHAT
keywords /* which variables */
>>WHEN
keywords /* which iterations & time steps */
>>WHERE
keywords /* in which blocks or planes */
.,,
CFX-4
Introductory Training (Dec 1999)
8 -15
CFX-4.3 V4.3.1H
How to get Convergence
? Under- relaxation:
can be used to try and obtain convergence,
if the transport equations are under-relaxed they
change to:
where URF =< 1.0 is the under-relaxation factor
and Φ0 is the value of the variable at the previous
iteration.
0)0.1(
ppunb
nb
nbp
p A
U R F
U R FSA
U R F
A ??? -??- ?
CFX-4
Introductory Training (Dec 1999)
8 -16
CFX-4.3 V4.3.1H
How to get Convergence
? Under- relaxation (cont.),
under-relaxing the variables makes the matrix
more diagonally dominant and thus easier to
solve,
often used to avoid divergence in the iterative
solution of strongly nonlinear equations (outer
iterations).
if convergence is poor,user may decrease the
default underrelaxation factors.
very small underrelaxation factors can slow down
the convergence significantly.
CFX-4
Introductory Training (Dec 1999)
8 -17
CFX-4.3 V4.3.1H
How to get Convergence
? False time-steps:
includes the pseudo time derivative so the
transport equations become:
generally useful in buoyant and multi-phase flows.
at the beginning of a calculation small false time-
step on the velocity variables will allow a sensible
solution to be obtained for other variables such as
temperature,The time-steps can then be
gradually increased,
Need to know about the time-scale for the
problem,this can be difficult.
0
punb
nb
nbpp t
VSA
t
VA ?????
???-??
??
?
?
?? ?
CFX-4
Introductory Training (Dec 1999)
8 -18
CFX-4.3 V4.3.1H
How to get Convergence
? Courant false time-steps,
avoids having to know the time scales of the
problem.
Calculates the time scale from the grid size and
local velocity field.
Different time-steps can be set for different
variable equations.
Achieved by setting a multiple of the local Courant
number time scale for each equation.
possible to have both false time-steps and under-
relaxation.
CFX-4
Introductory Training (Dec 1999)
8 -19
CFX-4.3 V4.3.1H
How to get Convergence
? Modify the inner iterations
using reduction factor:
ratio of residuals between
the last and first inner
iterations.
large reductions factors
(>1.0) indicate the linear
equation solver needs to be
improved.
linear equations can be
solved more accurately by
using the AMG solver,
increasing the maximum
number of sweeps and
decreasing the target
reduction factors.
sweeps
inner
iterationsu momen.
.,,,,,
.,,,,,
.,,,,,
CFX-4
Introductory Training (Dec 1999)
8 -20
CFX-4.3 V4.3.1H
How to get Convergence
? How to modify the
underrelaxation factors.
>>SOLVER DATA
>>UNDER RELAXATION FACTORS
U VELOCITY 0.65
V VELOCITY 0.65
W VELOCITY 0.65
PRESSURE 1.0
ENTHALPY 1.0
? How to modify the linear
equation solvers.
>>SOLVER DATA
>>EQUATION SOLVERS
U VELOCITY ‘STONE’
V VELOCITY ‘STONE’
W VELOCITY ‘STONE’
PRESSURE ‘ICCG’
ENTHALPY ‘AMG’
CFX-4
Introductory Training (Dec 1999)
8 -21
CFX-4.3 V4.3.1H
How to get Convergence
? Deferred Correction,
non-orthogonal cells in a turbulent problem may
cause convergence difficulties.
DEFERRED CORRECTION command can reduce,or
cut out,the cross-derivative diffusion term in the k
and ε equation.
For example:
>>SOLVER OPTIONS
…
>>DEFERRED CORRECTION
K START 1
K END 100
EPSILON START 1
EPSILON END 100
CFX-4
Introductory Training (Dec 1999)
8 -22
CFX-4.3 V4.3.1H
Convergence Difficulties
? If the residuals do not show convergence
check the command file
> fluid properties
> model data
> boundary conditions.
check user Fortran.
check the output file
> warning or error messages
> default values.
CFX-4
Introductory Training (Dec 1999)
8 -23
CFX-4.3 V4.3.1H
Convergence Difficulties
check the initial guess
> complex problems need more reasonable initial flow
field,so restart from simplified calculations.
check the linear solver performance
> average reduction factors
> SOLVER DEBUG PRINTING
> for poor linear solver performance,decrease the target
reduction factors and increase the maximum number of
sweeps.
CFX-4
Introductory Training (Dec 1999)
8 -24
CFX-4.3 V4.3.1H
Convergence Difficulties
check the grid quality
> orthogonality,grid stretch (<2 in general),etc.
> residuals in the whole field.
> to improve you may need to re-build the block structure
or re-mesh the geometry.
poor good
CFX-4
Introductory Training (Dec 1999)
8 -25
CFX-4.3 V4.3.1H
Convergence Difficulties
reduce the under-relaxation factors if the
residuals display unstable oscillations.
use AMG linear solver for complicated block
structures.
? See chapter 6 of CFX-4,Solver manual for
more details.
? Contact CFX Support line.