[ktgfeel]

Running 3D LES on FLUENT.

Clicked Data Sampling for Time Statistics options.
Now I see Resolved UV,UW,VW Reynolds Stresses on export quantities.

Q1. Why suddenly showing up after Time Statistics options??

Q2. Can I use this as the input for Acoustic Analogies?! Is it mean valued throughout the simulation?!

Q3. Do I just square the resolved x velocity to get Resolved UU Reynolds Stress? Is this correct?

Anyone experienced in dealing with Lighthill's Stress Tensor??

[LuckyTran]

Originally Posted by ktgfeel

Q1. Why suddenly showing up after Time Statistics options??

The Reynolds stresses are time-averaged quantities. Once you activate time statistics, many time-averaged mean and rms quantities suddenly become available (such as mean velocity, rms velocities).

Originally Posted by ktgfeel

Q2. Can I use this as the input for Acoustic Analogies?! Is it mean valued throughout the simulation?!

It is mean valued based on the sampling interval that you set (default is every time-step). And yes you can use it for a lot of stuff.

Originally Posted by ktgfeel

Q3. Do I just square the resolved x velocity to get Resolved UU Reynolds Stress? Is this correct?

If you do it to the rms velocities. Yes.

Sorry I am not familiar with Lighthill's Stress Tensor.

I am not sure why in the GUI the reynolds stresses are labeled with "Resolved". In the field variables they are not named that way but are simnply named UU UV and UW Reynolds stresses. Information on the meaning of "resolved" is somewhat obscure.

[ktgfeel]

Thank you very much for your reply. I know understand more. but still haven't reached my goal.

Fluent doesn't have any option on Time Staticstics to have set the length of time to calculate. It only has option to set sampling intervals.
However, what I am trying to get is somewhat like Instantaneous Reynolds Stress.

I have moved on to LES calculating method because having Reynolds Stress Model converge was almost impossible. ( Reynolds Stress Model,i.e. RSM, does calculate Reynolds stresses right away).

Q1. If I just multiply LES calculated(not rms) velocities, what would it mean?? Would it be something like instantaneous reynolds stresses??

Thank you again for your reply and careful explanation.

[ktgfeel]

Originally Posted by LuckyTran

The Reynolds stresses are time-averaged quantities. Once you activate time statistics, many time-averaged mean and rms quantities suddenly become available (such as mean velocity, rms velocities).

If you do it to the rms velocities. Yes.

Thank you very much for your reply. I know understand more. but still haven't reached my goal.

Fluent doesn't have any option on Time Staticstics to have set the length of time to calculate. It only has option to set sampling intervals.
However, what I am trying to get is somewhat like Instantaneous Reynolds Stress.

I have moved on to LES calculating method because having Reynolds Stress Model converge was almost impossible. ( Reynolds Stress Model,i.e. RSM, does calculate Reynolds stresses right away).

Q1. If I just multiply LES calculated(not rms) velocities, what would it mean?? Would it be something like instantaneous reynolds stresses??

Thank you again for your reply and careful explanation.

[Andrea1984]

Hi ktgfeel,

I know this is quite an old thread, I was just wondering if you or anybody else reading eventually got how the Resolved Reyonlds Stresses are calculated in Fluent when doing LES and collecting statistics.

My idea is that, for instance, the resolved uv stress should be defined as:

Resolved uv Reynolds stress = {u''v''}
where {} indicates ensemble average over time and

u''=[u]-{[u]}
v''=[v]-{[v]}
where [u] and [v] are the filtered (i.e. resolved) instantaneous velocity components, and {} again indicates time averaging.

Cheers
Andrea

[Andrea1984]

I have created custom field functions for the variables defined in my previous post and I am matching the Resolved Reynolds Stressed reported by Fluent under Unsteady Statistics.

I guess this answers LuckyTran's question about the label "resolved", since the Reynolds stressed are calculated based on the resolved velocity field and the contribution coming from the subgrid scales is neglected.

[LuckyTran]

Originally Posted by Andrea1984

I have created custom field functions for the variables defined in my previous post and I am matching the Resolved Reynolds Stressed reported by Fluent under Unsteady Statistics.

I guess this answers LuckyTran's question about the label "resolved", since the Reynolds stressed are calculated based on the resolved velocity field and the contribution coming from the subgrid scales is neglected.

Thanks I've always wondered but never got around to testing it myself. So it seems like Fluent needs to do a better job labeling internal variables and work on their documentation.

Originally Posted by Andrea1984

Hi ktgfeel,

I know this is quite an old thread, I was just wondering if you or anybody else reading eventually got how the Resolved Reyonlds Stresses are calculated in Fluent when doing LES and collecting statistics.

My idea is that, for instance, the resolved uv stress should be defined as:

Resolved uv Reynolds stress = {u''v''}
where {} indicates ensemble average over time and

u''=[u]-{[u]}
v''=[v]-{[v]}
where [u] and [v] are the filtered (i.e. resolved) instantaneous velocity components, and {} again indicates time averaging.

Cheers
Andrea

In most circles, this is the meaning of resolved Reynolds stress, calculated based on resolved/filtered velocity and does not include the contribution from the SGS model. It is good to know that the Fluent definition is the same. I appreciate the work you did to verify this.

[MarkelRF]

Originally Posted by Andrea1984

I have created custom field functions for the variables defined in my previous post and I am matching the Resolved Reynolds Stressed reported by Fluent under Unsteady Statistics.

I guess this answers LuckyTran's question about the label "resolved", since the Reynolds stressed are calculated based on the resolved velocity field and the contribution coming from the subgrid scales is neglected.

I have read posts from this thread and I am really interested in plotting these Resolved Reynolds Stresses. The thing is that I also have to plot {u"u"}, {v"v"} and {w"w"}.

u''=[u]-{[u]}
v''=[v]-{[v]}
w"= [w]-{[w]}
where [u], [v] and [w] are the filtered (i.e. resolved) instantaneous velocity components, and {} again indicates time averaging.

I am able to define the products between the brackets u"u" and so on, but I am not sure how to performe the time averaging via Custom Field Functions. Any clue on this would be highly appreciated.

[LuckyTran]

Originally Posted by MarkelRF

I am able to define the products between the brackets u"u" and so on, but I am not sure how to performe the time averaging via Custom Field Functions. Any clue on this would be highly appreciated.

I'm not sure why you need to do any custom field functions for the Reynolds stresses, these are already available if you enable data sampling. My guess is you have not yet found this super useful option. It's on the calculation tab in the GUI and in /solve/ in the TUI.

It's only doable with custom field functions unless you have v18. Before v18, custom field functions did not work in parallel (on serial) and you're probably not running LES on one core.

Once you have defined the custom field functions, there is an option to also collect statistics for the time-averaging.

But again, Fluent already does all this for you and you don't need to do anything except to click a button to enable "data sampling for time statistics"

The concerns mentioned above are that these are only the resolved parts of the Reynolds stresses based on filtered velocities. You need to then add the subgrid scale contribution.

[MarkelRF]

Originally Posted by LuckyTran

I'm not sure why you need to do any custom field functions for the Reynolds stresses, these are already available if you enable data sampling. My guess is you have not yet found this super useful option. It's on the calculation tab in the GUI and in /solve/ in the TUI.

But again, Fluent already does all this for you and you don't need to do anything except to click a button to enable "data sampling for time statistics"

The concerns mentioned above are that these are only the resolved parts of the Reynolds stresses based on filtered velocities. You need to then add the subgrid scale contribution.

Thanks for your quick reply LuckyTran. I have been looking for what you said but I am not able to find anything. I have indeed run the LES simulation with 4 cores and my fluent version is 17.1.

I have already applied the sampling statistics and I already have some available data at unsteady statistics subwidnow, the problem is that the resolved reynolds stresses that appear in my fluent window are uv, uw, vw; but no more. I still need uu, vv and ww.

On the other hand, how would you add the contribution of the SGS to the reynolds stresses?

Finally, and this is out of topic, when I open my unsteady statistics window, I do not get to see RSM velocities, but RSME (Root Mean Square Error). Do you know why this could be?

Yours,

MarkelRF

[LuckyTran]

The RMS of the fluctuating velocity RMS of u' is the RMSE of u. It is a common mistake and even early versions of fluent incorrectly labeled the velocities as RMS when the calculation was actually an RMSE calculation.

You can get the uu stress from the square of the RMSE of u

The subgrid scale reynolds stresses are non-trivial, and for this you do need custom field functions or possibly udf's. The short answer is that you need to be familiar with the particular sgs model and use its definition (i.e. Boussinesq hypothesis). You'll need 6 functions, one for each Reynolds stress.

[roi247]

35.4. Alphabetical Listing of Field Variables and Their Definitions


In the Fluent User Guide 19.2 (p. 2497), it seems they mean to involve the SGS now.