Restarting simulation from instantenous & average velocity field files
Hello Paul & Stefan, I am running a DNS simulation of backward facing simulation (BFS) at Re=3000. The Reynolds # is based on step height (h=1) & free stream velocity (ux = 1.0). The BFS geometry has 5184 elements and lx1=ly1=lz1=8. After running for about 10 flow throughs, we started computing averages. I want to restart simulation after 2 flow throughs, but I have 2 field files one for instantaneous (blah.fld01) and the other for average (avgblah.fld01). To continue computing averages I need to read in both the velocity field files. Is there an option to restart from 2 field files? Or how else can I do this? I will also require this option later for computing "rms" velocities. Thanks, Harish.
Harish, you can compute mean and variances based on the averages we computed in avg_all(). The averages E(X), E(X*X), E(X*Y) are stored in the common blocks defined in AVG (make sure you set lxa in SIZE correct). The averages are dumped into different files (I/O freq can be adjusted with param(68): *.avg -- E(X) *.rms -- E(X*X) *.rm2 -- E(X*Y) If you have want to compute the mean and variance of multiple files, run Nek in postprocessing mode and read the *.avg files and the *.rms files using load_fld() and sum up the averages for every variable (you need some temporary arrays for that). Then compute the variance as var(X) := E(X^X) - E(X)*E(X) Finally just dump the final arrays containing your mean and/or variances. hth, Stefan On Apr 30, 2010, at 2:16 PM, [email protected] wrote:
Hello Paul & Stefan,
I am running a DNS simulation of backward facing simulation (BFS) at Re=3000. The Reynolds # is based on step height (h=1) & free stream velocity (ux = 1.0). The BFS geometry has 5184 elements and lx1=ly1=lz1=8.
After running for about 10 flow throughs, we started computing averages. I want to restart simulation after 2 flow throughs, but I have 2 field files one for instantaneous (blah.fld01) and the other for average (avgblah.fld01). To continue computing averages I need to read in both the velocity field files. Is there an option to restart from 2 field files? Or how else can I do this? I will also require this option later for computing "rms" velocities.
Thanks,
Harish.
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
Hi Paul & Stefan, Thanks for your replies. I read through the routine avg_all() and I am using that to compute averages. I am also using userchk subroutine in turbChannel case as a guideline. I am sorry for not being very clear about my simulations. For the DNS simulation I ran it for 10 flow throughs till quasi-stationary state is achieved. Then compute averages over 5 more flow throughs and finally compute fluctuations over 10 more flow throughs. Right now I am computing averages by setting up the simulation for all the 5 flow throughs at one time. At the end of 5 flow throughs I will end up with 2 fld* files. One for instantaneous (blah.fld*) and other for averages (avgblah.fld*). To continue with calculating fluctuations over the next 10 flow throughs. I need to read in the instantaneous and average field values to calculate fluctuations. Is there is possible to do this with the RESTART OPTIONS. Something like this, 2 PRESOLVE/RESTART OPTIONS ***** blah.fld01 X U P avgblah.fld01 Uavg Pavg I think there is one way around this. To calculate fluctuations, start from fresh averaging for next 10 flow throughs and not use the averages calculated from previous 5 flow throughs. I am not sure if this is the right way to go about it. Thanks for your help. I will test some cases out and will post any questions I have. Regards, Harish. On Fri, April 30, 2010 7:37 am, [email protected] wrote:
Harish,
you can compute mean and variances based on the averages we computed in avg_all(). The averages E(X), E(X*X), E(X*Y) are stored in the common blocks defined in AVG (make sure you set lxa in SIZE correct).
The averages are dumped into different files (I/O freq can be adjusted with param(68): *.avg -- E(X) *.rms -- E(X*X) *.rm2 -- E(X*Y)
If you have want to compute the mean and variance of multiple files, run Nek in postprocessing mode and read the *.avg files and the *.rms files using load_fld() and sum up the averages for every variable (you need some temporary arrays for that).
Then compute the variance as
var(X) := E(X^X) - E(X)*E(X)
Finally just dump the final arrays containing your mean and/or variances.
hth, Stefan
On Apr 30, 2010, at 2:16 PM, [email protected] wrote:
Hello Paul & Stefan,
I am running a DNS simulation of backward facing simulation (BFS) at Re=3000. The Reynolds # is based on step height (h=1) & free stream velocity (ux = 1.0). The BFS geometry has 5184 elements and lx1=ly1=lz1=8.
After running for about 10 flow throughs, we started computing averages. I want to restart simulation after 2 flow throughs, but I have 2 field files one for instantaneous (blah.fld01) and the other for average (avgblah.fld01). To continue computing averages I need to read in both the velocity field files. Is there an option to restart from 2 field files? Or how else can I do this? I will also require this option later for computing "rms" velocities.
Thanks,
Harish.
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
If you're trying to compute the rms quantities, you can do these directly from the data generated by avg_all, which produces <u> <v> <w> and <u^2> <v^2> <w^2> <uv> <vw> etc. Given these, you have <u'2> = <u^2> - <u>^2 etc., so no need to rerun if you're after those statistics. Paul On Sat, 1 May 2010, [email protected] wrote:
Hi Paul & Stefan,
Thanks for your replies.
I read through the routine avg_all() and I am using that to compute averages. I am also using userchk subroutine in turbChannel case as a guideline.
I am sorry for not being very clear about my simulations. For the DNS simulation I ran it for 10 flow throughs till quasi-stationary state is achieved. Then compute averages over 5 more flow throughs and finally compute fluctuations over 10 more flow throughs.
Right now I am computing averages by setting up the simulation for all the 5 flow throughs at one time. At the end of 5 flow throughs I will end up with 2 fld* files. One for instantaneous (blah.fld*) and other for averages (avgblah.fld*).
To continue with calculating fluctuations over the next 10 flow throughs. I need to read in the instantaneous and average field values to calculate fluctuations. Is there is possible to do this with the RESTART OPTIONS. Something like this,
2 PRESOLVE/RESTART OPTIONS ***** blah.fld01 X U P avgblah.fld01 Uavg Pavg
I think there is one way around this. To calculate fluctuations, start from fresh averaging for next 10 flow throughs and not use the averages calculated from previous 5 flow throughs. I am not sure if this is the right way to go about it.
Thanks for your help. I will test some cases out and will post any questions I have.
Regards,
Harish.
On Fri, April 30, 2010 7:37 am, [email protected] wrote:
Harish,
you can compute mean and variances based on the averages we computed in avg_all(). The averages E(X), E(X*X), E(X*Y) are stored in the common blocks defined in AVG (make sure you set lxa in SIZE correct).
The averages are dumped into different files (I/O freq can be adjusted with param(68): *.avg -- E(X) *.rms -- E(X*X) *.rm2 -- E(X*Y)
If you have want to compute the mean and variance of multiple files, run Nek in postprocessing mode and read the *.avg files and the *.rms files using load_fld() and sum up the averages for every variable (you need some temporary arrays for that).
Then compute the variance as
var(X) := E(X^X) - E(X)*E(X)
Finally just dump the final arrays containing your mean and/or variances.
hth, Stefan
On Apr 30, 2010, at 2:16 PM, [email protected] wrote:
Hello Paul & Stefan,
I am running a DNS simulation of backward facing simulation (BFS) at Re=3000. The Reynolds # is based on step height (h=1) & free stream velocity (ux = 1.0). The BFS geometry has 5184 elements and lx1=ly1=lz1=8.
After running for about 10 flow throughs, we started computing averages. I want to restart simulation after 2 flow throughs, but I have 2 field files one for instantaneous (blah.fld01) and the other for average (avgblah.fld01). To continue computing averages I need to read in both the velocity field files. Is there an option to restart from 2 field files? Or how else can I do this? I will also require this option later for computing "rms" velocities.
Thanks,
Harish.
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
Hi Paul, Thanks for your info. on how to find the rms quantities. This will certainly save me a lot of computing time. Regards, Harish. On Sat, 1 May 2010 [email protected] wrote:
If you're trying to compute the rms quantities, you can do these directly from the data generated by avg_all, which produces
<u> <v> <w>
and <u^2> <v^2> <w^2> <uv> <vw> etc.
Given these, you have
<u'2> = <u^2> - <u>^2
etc., so no need to rerun if you're after those statistics.
Paul
On Sat, 1 May 2010, [email protected] wrote:
Hi Paul & Stefan,
Thanks for your replies.
I read through the routine avg_all() and I am using that to compute averages. I am also using userchk subroutine in turbChannel case as a guideline.
I am sorry for not being very clear about my simulations. For the DNS simulation I ran it for 10 flow throughs till quasi-stationary state is achieved. Then compute averages over 5 more flow throughs and finally compute fluctuations over 10 more flow throughs.
Right now I am computing averages by setting up the simulation for all the 5 flow throughs at one time. At the end of 5 flow throughs I will end up with 2 fld* files. One for instantaneous (blah.fld*) and other for averages (avgblah.fld*).
To continue with calculating fluctuations over the next 10 flow throughs. I need to read in the instantaneous and average field values to calculate fluctuations. Is there is possible to do this with the RESTART OPTIONS. Something like this,
2 PRESOLVE/RESTART OPTIONS ***** blah.fld01 X U P avgblah.fld01 Uavg Pavg
I think there is one way around this. To calculate fluctuations, start from fresh averaging for next 10 flow throughs and not use the averages calculated from previous 5 flow throughs. I am not sure if this is the right way to go about it.
Thanks for your help. I will test some cases out and will post any questions I have.
Regards,
Harish.
On Fri, April 30, 2010 7:37 am, [email protected] wrote:
Harish,
you can compute mean and variances based on the averages we computed in avg_all(). The averages E(X), E(X*X), E(X*Y) are stored in the common blocks defined in AVG (make sure you set lxa in SIZE correct).
The averages are dumped into different files (I/O freq can be adjusted with param(68): *.avg -- E(X) *.rms -- E(X*X) *.rm2 -- E(X*Y)
If you have want to compute the mean and variance of multiple files, run Nek in postprocessing mode and read the *.avg files and the *.rms files using load_fld() and sum up the averages for every variable (you need some temporary arrays for that).
Then compute the variance as
var(X) := E(X^X) - E(X)*E(X)
Finally just dump the final arrays containing your mean and/or variances.
hth, Stefan
On Apr 30, 2010, at 2:16 PM, [email protected] wrote:
Hello Paul & Stefan,
I am running a DNS simulation of backward facing simulation (BFS) at Re=3000. The Reynolds # is based on step height (h=1) & free stream velocity (ux = 1.0). The BFS geometry has 5184 elements and lx1=ly1=lz1=8.
After running for about 10 flow throughs, we started computing averages. I want to restart simulation after 2 flow throughs, but I have 2 field files one for instantaneous (blah.fld01) and the other for average (avgblah.fld01). To continue computing averages I need to read in both the velocity field files. Is there an option to restart from 2 field files? Or how else can I do this? I will also require this option later for computing "rms" velocities.
Thanks,
Harish.
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
Hi Harish, 2 PRESOLVE/RESTART OPTIONS ***** blah.fld01 X blah.fld02 U P will take geometry (say) from .fld01 and velocity/pressure from fld02, as well as time from the last file listed. In your case, why not just restart, and then subsequently average your avg files ? We have some small codes for doing this. Paul On Fri, 30 Apr 2010, [email protected] wrote:
Hello Paul & Stefan,
I am running a DNS simulation of backward facing simulation (BFS) at Re=3000. The Reynolds # is based on step height (h=1) & free stream velocity (ux = 1.0). The BFS geometry has 5184 elements and lx1=ly1=lz1=8.
After running for about 10 flow throughs, we started computing averages. I want to restart simulation after 2 flow throughs, but I have 2 field files one for instantaneous (blah.fld01) and the other for average (avgblah.fld01). To continue computing averages I need to read in both the velocity field files. Is there an option to restart from 2 field files? Or how else can I do this? I will also require this option later for computing "rms" velocities.
Thanks,
Harish.
_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
participants (1)
-
nek5000-users＠lists.mcs.anl.gov