LES of channel flow asymetric u2prime

ztdep · September 26, 2023, 20:07

We employed the dynamic K-LES model to simulate channel flow, operating at a Reynolds number 180 As depicted in the figure, the results have been averaged over 200 flow-through times. However, the u2prime profile exhibits an asymmetric pattern.
Any suggestions or insights to address this anomaly would be greatly appreciated.

Tobermory · September 27, 2023, 10:55

Just run for longer. Second order statistics require longer averaging times ...

Tobermory · September 27, 2023, 11:11

If you want to estimate the averaging time, take a look at Lumley & Panofsky where you'll find a simple method for estimating the required time based on the integral scale of the turbulence. I don't know how long your domain is, but assuming it's around 10 times the channel halfwidth, then your 5% error is roughly what I'd expect after 200 flow throughs. The error is proportional to the square root of the number of flow throughs ... so you'll need to run about 3x longer to drop the error from 5% to 3% etc.

By the way, a smarter approach is to average across the top and bottom halves of your domain, since it's symmetric (clearly you need to average from the wall in both halves ...!). That will halve your run time!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
LES of turbulent channel flow (overprediction)	nathanricks	FLUENT	1	March 3, 2024 21:47
unable to run dynamic mesh(6dof) and wave UDF	shedo	Fluent UDF and Scheme Programming	0	July 1, 2022 17:22
LES of turbulent channel flows	cedric_duprat	OpenFOAM Running, Solving & CFD	220	November 18, 2019 04:19
Velocity curve in channel flow by pisoFoam and LES	mxylondon	OpenFOAM Verification & Validation	9	June 21, 2015 07:28
Fluent LES Channel Flow	santosh	FLUENT	0	January 7, 2010 21:11

September 27, 2023, 10:55		#2
Tobermory Senior Member Join Date: Apr 2020 Location: UK Posts: 668 Rep Power: 14	Just run for longer. Second order statistics require longer averaging times ...

September 27, 2023, 11:11		#3
Tobermory Senior Member Join Date: Apr 2020 Location: UK Posts: 668 Rep Power: 14	If you want to estimate the averaging time, take a look at Lumley & Panofsky where you'll find a simple method for estimating the required time based on the integral scale of the turbulence. I don't know how long your domain is, but assuming it's around 10 times the channel halfwidth, then your 5% error is roughly what I'd expect after 200 flow throughs. The error is proportional to the square root of the number of flow throughs ... so you'll need to run about 3x longer to drop the error from 5% to 3% etc. By the way, a smarter approach is to average across the top and bottom halves of your domain, since it's symmetric (clearly you need to average from the wall in both halves ...!). That will halve your run time!