{CFL# | TIMESTEP} { \ [MODE 1] [CFL | SECONDS] cfl1 [zone_selector [ityp]] | \ [MODE 2] [CFL | SECONDS] INCREMENT cfl1 cflmax cflfac inccfl istart \ [zone_selector [ityp]] | \ MODE 3 [DT cfl1] [CFLMIN cflmin] [CFLMAX cflmax] [ITIME itime] \ [zone_selector]} |
{CFL# | TIMESTEP} { \ [MODE 1] [CFL | SECONDS] cfl1 [zone_selector] | \ [MODE 2] [CFL | SECONDS] INCREMENT cfl1 cflmax cflfac inccfl istart \ [zone_selector] | \ {AUTOMATIC | MODE 4} \ [CFLSTART cfl1] [CFLMIN cflmin] [CFLMAX cflmax] \ [INCREASE incrfac] [DECREASE decrfac] [FREQUENCY freq] \ [MAX_RATIO maxratio] [TRY_INCREASE tryincr] \ [zone_selector]} |
This keyword allows the user to specify the CFL number or time step for all zones, or on a zone-by-zone basis. All keywords and parameters must be on one line in the data file. If this keyword is not used, a constant CFL number is used, with a value of 1.3 for structured grids and 10.0 for unstructured grids. For unstructured grids, a larger CFL value can usually be used than for structured grids. Suggested values for unstructured grids are 10 at the start of a calculation, and as high as 50 for later runs.
For structured grids there are three different options, as specified by
the MODE value.
Note that the MODE keyword is optional for modes 1 and 2, but
required for mode 3.
(Modes 1 and 2 are distinguished in the code by whether or not the word
INCREMENT is present.)
{CFL# | TIMESTEP} [MODE 1] [CFL | SECONDS] cfl1 [zone_selector [ityp]] |
In this mode, the CFL number or time step is specified directly, by the
input value cfl1.
CFL | A CFL number is being specified.
This is the default, and should be used for steady flow
problems.
| ||
---|---|---|---|
SECONDS | The time step is being specified in seconds. This option is generally used for unsteady flow problems. |
With single-zone grids, when a CFL number is being specified the value
of ityp may be set to 1 to indicate that a global time step
(i.e., constant in space) should be used, equal to the minimum value in
the zone.
This allows the time step to be determined through a CFL number for
unsteady flow problems.
When a CFL number is specified, and ityp = 0 or is omitted,
a local time step (i.e., varying in space) will be used.
{CFL# | TIMESTEP} [MODE 2] [CFL | SECONDS] INCREMENT \ cfl1 cflmax cflfac inccfl istart [zone_selector [ityp]] |
In this mode, the CFL number or time step will be gradually increased as
the calculation proceeds.
The CFL and SECONDS options have the same meaning as
for mode 1, again with CFL as the default.
The increase in the CFL number or time step is controlled by the
following input values:
cfl1 | CFL number or time step for iteration istart | ||
---|---|---|---|
cflmax | Maximum CFL number or time step allowed | ||
cflfac | Factor which multiplies the CFL number or time step every inccfl iterations | ||
inccfl | Iteration increment at which the CFL number or time step is multiplied by cflfac. The multiplication is done whenever mod(iteration,inccfl) = 1. | ||
istart | Starting iteration in the increment calculation. Note that the iteration values are cumulative, starting from the initial run. I.e., to gradually increase the CFL number at the beginning of a restart case, istart should be set to the value of the iteration in the existing .cfl file, from the previous run. |
As in mode 1, when a CFL number is being specified for a single-zone grid, the value of ityp may be set to 1 to indicate that a global time step (i.e., constant in space) should be used, equal to the minimum value in the zone.
If mode 2 is used in a staged solution, it should only be used at the
start of the run, and SET CFL shouldn't be specified in the
SOLVER-STAGES block for the
first stage.
{CFL# | TIMESTEP} MODE 3 [DT cfl1] [CFLMIN cflmin] [CFLMAX cflmax] \ [ITIME itime] [zone_selector] |
This mode uses the time step calculation procedure originally used in
the OVERFLOW code.
itime | A flag indicating the type of time step to be used
The default value is 1. | ||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
cfl1 | Time step parameter.
For itime = 0 (i.e., time-accurate cases), this
is the time step in seconds nondimensionalized by
L_{r} / a_{r}, where
L_{r} is the grid reference length and
a_{r} is the freestream speed of sound.
For itime = 1 and 2, this is a CFL number,
before scaling by the local metric Jacobian.
For itime = 3, cfl1 is not used.
The default value is 0.5.
| ||||||||||||||||||
cflmin, cflmax | For itime = 3, the CFL number is set to
cflmax, and cflmin is not used.
The CFL number is defined using the sum of the maximum
eigenvalues in each coordinate direction.
For other itime values, cflmin and cflmax are used to limit the CFL number to values within the specified range. A value of 0.0 indicates no limiting. If either cflmin or cflmax is negative, the absolute values are used, and the CFL number is defined using the method of Gnoffo, with a viscous correction by Tannehill. If both cflmin and cflmax are non-negative, the usual one-dimensional (inviscid) CFL number definition using the maximum eigenvalue is used. The default values are both 0.0. |
If mode 3 is used, it must be used in all zones, but the values of cfl1, cflmin, etc., may vary from zone to zone.
See Also: CROSSFLOW, NEWTON, TEMPORAL
There are also three options for specifying the CFL number or time step
for unstructured grids.
Note that the MODE keyword is optional for the first two
options, but the word AUTOMATIC (or optionally MODE 4),
is required for the third option.
{CFL# | TIMESTEP} [MODE 1] [CFL | SECONDS] cfl1 [zone_selector] |
This mode is essentially the same as the MODE
1 option for structured grids, described above.
However, with unstructured grids the ityp input option, which may
be used with single-zone structured grids when a CFL number is being
specified to indicate that a global time step (i.e., constant in space)
should be used, is not available.
In addition, for unstructured grids with local time stepping, the size
of the physical time step will automatically be limited to the largest
time step in the grid that corresponds to a CFL number of one.
Of course, due to variations in grid spacing and flow conditions, the
CFL number at other locations may be much higher.
{CFL# | TIMESTEP} [MODE 2] [CFL | SECONDS] INCREMENT \ cfl1 cflmax cflfac inccfl istart [zone_selector] |
This mode is essentially the same as the MODE
2 option for structured grids, described above.
Like MODE 1, with unstructured grids the ityp input
option is not available, and with local time stepping, the size of the
physical time step will automatically be limited to the largest time
step in the grid that corresponds to a CFL number of one.
{CFL# | TIMESTEP} {AUTOMATIC | MODE 4} \ [CFLSTART cfl1] [CFLMIN cflmin] [CFLMAX cflmax] \ [INCREASE incrfac] [DECREASE decrfac] [FREQUENCY freq] \ [MAX_RATIO maxratio] [TRY_INCREASE tryincr] [zone_selector]} |
With this mode, the CFL number is automatically decreased or increased based on the convergence behavior of the solution. Note that like all the other modes, the keyword may be specified as CFL# or TIMESTEP, but this mode specifies only a CFL number, not a physical time step.
The input values are:
cfl1 | Starting CFL number. The default is 10.0. | ||
---|---|---|---|
cflmin | Minimum CFL number that will be used. The default is 0.1. | ||
cflmax | Maximum CFL number that will be used. The default is 100×cfl1. | ||
incrfac | Factor by which the CFL number will be increased in a non-diverging solution. The default is 2.0. | ||
decrfac | Factor by which the CFL number will be decreased in a diverging solution. The default is 10.0. | ||
freq | Iteration frequency for checking whether or not the CFL number should be increased in a non-diverging solution. This is referenced to the last iteration in which the CFL number was changed (increased or decreased). The default is 10. | ||
maxratio | Maximum allowed ratio of CFL number among zones. The default is 5. | ||
tryincr | Interval for increasing the CFL number in a solution that is neither converging or diverging. The default is 3. |
The solution is checked for divergence every iteration, and if it is diverging, the CFL number is decreased by the factor decrfac, down to a minimum of cflmin.
In a non-diverging solution, the solution is checked every freq iterations to determine whether or not the CFL number should be increased. If the solution is converging, the CFL number is increased by the factor incrfac. If the solution is neither converging or diverging, the CFL number is also increased by the factor incrfac, but only every tryincr times this check is made (i.e., every freq×tryincr iterations). The CFL number in a particular zone is limited to the value of cflmax for that zone, or 5 times the smallest CFL number for all the other zones, whichever is lower.
The ΔQ limiter (keyword DQ LIMITER) must be on (which is the default) when CFL AUTOMATIC is used.
See Also: NEWTON, TEMPORAL, DQ LIMITER
Examples
Set the CFL number to 1.5 in all zones.
CFL# CFL 1.5
Set the CFL number to 1.5 in zones 1 and 5, and 0.7 in zones 2, 3, and 4.
CFL# CFL 1.5 ZONE 1,5 CFL# CFL 0.7 ZONE 2:4
The following example sets the CFL number to 1.0, computes the corresponding time step at every point in the grid, finds the minimum of those time steps, and resets the time step at every point to that minimum value. This allows the time step to be set for an unsteady flow problem by specifying a CFL number. Note that even though this may only be done for single-zone grids, the syntax requires that the zone be explicitly specified.
CFL# CFL 1.0 ZONE 1 1
Set the time step equal to 1×10^{−6} seconds at every point.
CFL# SECONDS 0.000001
The next example sets the CFL number to 0.5 for the first 500 iterations, then increases it by a factor of 1.1 every 100 iterations. The maximum value allowed will be 2.0, so the actual final CFL number will be some value just below 2.0. (For this example, it's 1.899.)
CFL# CFL INCREMENT 0.5 2.0 1.1 100 501
The remaining examples use the mode 3 time step calculation procedure, only available for structured grids. The first example sets the time step parameter to 0.1, and defaults the remaining input parameters, causing the actual time step to be scaled by the local metric Jacobian with the 0.005 fudge factor, with no minimum or maximum value.
CFL# MODE 3 DT 0.1
In this example the mode 3 time step calculation procedure is used to set the time step parameter to 0.5. The actual time step will be scaled by the local metric Jacobian with the 0.005 fudge factor, with minimum and maximum CFL numbers of 1.0 and 5.0. Since the minimum and maximum values are specified as positive numbers, the standard one-dimensional (inviscid) CFL number definition using the maximum eigenvalue is used.
CFL# MODE 3 DT 0.5 CFLMIN 1.0 CFLMAX 5.0
The next example uses the mode 3 time step calculation procedure to set the time step parameter to 0.05. The actual time step will be scaled by the local metric Jacobian without the 0.005 fudge factor, with minimum and maximum CFL numbers of 1.0 and 5.0. Since the minimum and maximum values are specified as negative numbers, the CFL definition of Gnoffo is used, with a viscous correction by Tannehill.
CFL# MODE 3 DT 0.05 ITIME 2 CFLMIN -1.0 CFLMAX -5.0
Use the mode 3 time step calculation procedure to set the CFL number to a constant value of 1.3.
CFL# MODE 3 ITIME 3 CFLMAX 1.3
Use the mode 3 time step calculation procedure for an unsteady case, with the nondimensional time step set to a constant value of 0.05.
CFL# MODE 3 ITIME 0 DT 0.05
The following example uses the mode 3 time step calculation procedure for an unsteady case, again with a specified nondimensional time step of 0.05. In this case, however, if the specified time step results in a CFL number at some point in the flow field less than 0.25 or greater than 5.0, the actual time step throughout the flow field will be reset to the corresponding value.
CFL# MODE 3 ITIME 0 DT 0.05 CFLMIN 0.25 CFLMAX 5.0
Last updated 1 Apr 2016