[mona.li]

Dear all,

I am doing a transient simulation with a need to get the time averaged velocity field using the "arithmetic average" set in CFX-Pre.

But, when I check the time average results, I have found that for the nearly unchanged transient velocity profile which was extracted far upstream from the separation bubble and thus remains almost steady there (which has been checked from the transient data), its resulted time-averaged profile differ greatly from it.

In particular, the time-averaged profile oscillates a lot and is further more discontinuous along the normal direction (y), while the transient one is smooth. These can be found in the figures here called "u_trnavg changes" and "u_unchange".

u_trnavg_changes.jpg

u_unchange.jpg

From these figures you can also find that the issue became more and more worse with the increase of total time steps!!!

To find the reason, I tried to calculate the time average of this field by extracting the transient ".trn" results to matlab and get the mean of them. Unlike the above, this time the time-averaged profile perfectly matches the nearly steady transient one as early expected!!!

This really confused me and any help will be appreciated!!!

Thank you in advance!

Best Regards,
Mona

[ghorrocks]

Please post images directly into the forum, do not use third party sites. Here is how: https://www.cfd-online.com/Wiki/Ansy...n_the_forum.3F

I have no idea why you are getting these weird results. You are going to have to analyse in detail what you are doing and what could be happening. We cannot do this over the forum as you will need to look at exactly how you have set it up.

You may have found a bug in the transient averaging, or you may have not set it up correctly. You will have to figure it out.

If you post your output file we can try to help, but the chances of us finding the cause are slim.

[mona.li]

Originally Posted by ghorrocks

Please post images directly into the forum, do not use third party sites. Here is how: https://www.cfd-online.com/Wiki/Ansy...n_the_forum.3F

I have no idea why you are getting these weird results. You are going to have to analyse in detail what you are doing and what could be happening. We cannot do this over the forum as you will need to look at exactly how you have set it up.

You may have found a bug in the transient averaging, or you may have not set it up correctly. You will have to figure it out.

If you post your output file we can try to help, but the chances of us finding the cause are slim.

Thank you very much for your kindly suggestion and sorry for wrongly posting the pictures.

The figure below shows my setting on the "Trn Stats" part, which seems to be the only related part with the time-averaging in CFX considering that the time-averaging here seems to be a purely arithmetic process in theory which means it is not affected by how the model would be set.

Trn_Stats_setting.jpg

I don't know whether this thought is correct and if there is something wrong with my setting？

Besides, I hope to know if there is any way to know the algorithm of this function in CFX-solver?

Thank you very much in advance and any reply will be appreciated!

Best regards,
Mona

[ghorrocks]

It should be just a time weighted average, so a very simple function.

The best way to work out if it is behaving as expected is to give it some test cases and see if it does what you expect. On a simple test case (like flow in a pipe or something) does it give the expected results if nothing changes with time? If you linearly ramp it over time? Do a step change?

[mona.li]

Originally Posted by ghorrocks

It should be just a time weighted average, so a very simple function.

The best way to work out if it is behaving as expected is to give it some test cases and see if it does what you expect. On a simple test case (like flow in a pipe or something) does it give the expected results if nothing changes with time? If you linearly ramp it over time? Do a step change?

Thank you very much for the kindly reply.

My time step has never been changed during the whole statistic process.

I will next do some tests according to your suggestion.

Best regards,
Mona.