CFL evaluation for LES simulation

julien_g · January 13, 2024, 05:16

Hello,
I'm performing my first LES simulation: the flow inside a wavy channel.
I have 3 questions:

1. The CFL value should be less than 1: this refers to an average value of all domain, or it should be the maximum value (all cells of the domain should have CFL<=1)?

2. At the end of the simulation, when i do the time averaging of the results, i should consider a timespan equal to one or a multiple of the characteristic period of the flow? (The flow should exhibit a periodic oscillatory behaviour.)

3. When i plot a monitor in one point of the flow (velocity vs simulation time), it does show a sine-like shape, but the amplitudes are varying too much - it doesn't look like a stable pattern. I don't know what i'm doing wrong... Could this be caused by a too large time step?

It seems that I would need a very small time step size, this would make the simulation run for several days. I'm wondering if i could reduce the total duration.

Thank you.

FMDenaro · January 13, 2024, 06:06

Quote:

Originally Posted by julien_g

Hello,
I'm performing my first LES simulation: the flow inside a wavy channel.
I have 3 questions:

1. The CFL value should be less than 1: this refers to an average value of all domain, or it should be the maximum value (all cells of the domain should have CFL<=1)?

2. At the end of the simulation, when i do the time averaging of the results, i should consider a timespan equal to one or a multiple of the characteristic period of the flow? (The flow should exhibit a periodic oscillatory behaviour.)

3. When i plot a monitor in one point of the flow (velocity vs simulation time), it does show a sine-like shape, but the amplitudes are varying too much - it doesn't look like a stable pattern. I don't know what i'm doing wrong... Could this be caused by a too large time step?

It seems that I would need a very small time step size, this would make the simulation run for several days. I'm wondering if i could reduce the total duration.

Thank you.

1) you should ensure that in the cell where the happens the worst velocity and mesh size condition

2) sample the data at each turnover time, if you are working in non-dimensional variables, that can be done at each time-unit.
3) You should monitor the total (volume-averaged) kinetic energy in time to check when the arbitrary initial condition are disregarded and the solution is physically correlated.

LuckyTran · January 13, 2024, 16:06

Just to reaffirm and maybe say it in a slightly different language.

1) The maximum of all cells need to be less than 1. So yes, your dt will be very small if you have even 1 cell that is very small.
2) Average over multiples (i.e. 10x) the temporal waviness (approximately the wave-pitch / bulk velocity)
3) Turbulent flows should look very messy. Periodicity is only at a single specific spatialwavenumber/frequency, that signal at one frequency will interact with all the other non-periodic content and produce very messy timetraces.

January 13, 2024, 05:16	CFL evaluation for LES simulation	#1
julien_g New Member Join Date: Feb 2023 Location: France Posts: 3 Rep Power: 3	Hello, I'm performing my first LES simulation: the flow inside a wavy channel. I have 3 questions: 1. The CFL value should be less than 1: this refers to an average value of all domain, or it should be the maximum value (all cells of the domain should have CFL<=1)? 2. At the end of the simulation, when i do the time averaging of the results, i should consider a timespan equal to one or a multiple of the characteristic period of the flow? (The flow should exhibit a periodic oscillatory behaviour.) 3. When i plot a monitor in one point of the flow (velocity vs simulation time), it does show a sine-like shape, but the amplitudes are varying too much - it doesn't look like a stable pattern. I don't know what i'm doing wrong... Could this be caused by a too large time step? It seems that I would need a very small time step size, this would make the simulation run for several days. I'm wondering if i could reduce the total duration. Thank you.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Using results from simulation A in simulation B	Hafssa	FLUENT	2	October 9, 2022 03:19
SU2 NACA0012 Transitional flow simulation Convergence Issues	morgJ	SU2	0	July 21, 2022 07:42
Moving mesh simulation with mesh changes inside the solver	cgoessni	OpenFOAM Programming & Development	2	January 24, 2022 09:14
restarting CFDEM simulation: OpenFOAM coupled with Liggghts	atul1018	Main CFD Forum	0	December 13, 2021 04:51
Control simulation to apply different fields with chtMultiRegionFoam	jmdf	OpenFOAM Running, Solving & CFD	0	February 29, 2016 07:05

January 13, 2024, 16:06		#3
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,674 Rep Power: 65	Just to reaffirm and maybe say it in a slightly different language. 1) The maximum of all cells need to be less than 1. So yes, your dt will be very small if you have even 1 cell that is very small. 2) Average over multiples (i.e. 10x) the temporal waviness (approximately the wave-pitch / bulk velocity) 3) Turbulent flows should look very messy. Periodicity is only at a single specific spatialwavenumber/frequency, that signal at one frequency will interact with all the other non-periodic content and produce very messy timetraces.