CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Do I need Convergence Monitor for Unsteady Simulations? (http://www.cfd-online.com/Forums/main/106552-do-i-need-convergence-monitor-unsteady-simulations.html)

1/153 August 31, 2012 16:57

Do I need Convergence Monitor for Unsteady Simulations?
 
Dear All,

I saw a few posts on this forum that those people are monitoring the residuals or what for their unsteady simulations to see it converging or not. My question is: is it necessary? What are they monitoring? How to judge it converges or not?

My case is an unsteady RANS simulation around a bluff body with vortex shedding and moderate to high Re.

Once my boundary condition is right, the seatdy simulation results are good. Do I still need to monitor the convergence?

Thanks for the help.

1/153 dw

FMDenaro August 31, 2012 17:29

Quote:

Originally Posted by 1/153 (Post 379775)
Dear All,

I saw a few posts on this forum that those people are monitoring the residuals or what for their unsteady simulations to see it converging or not. My question is: is it necessary? What are they monitoring? How to judge it converges or not?

My case is an unsteady RANS simulation around a bluff body with vortex shedding and moderate to high Re.

Once my boundary condition is right, the seatdy simulation results are good. Do I still need to monitor the convergence?

Thanks for the help.

1/153 dw


In my opinion, controlling the unsteady simulation by monitoring the time derivatives can be useful.
- If the case is steady (laminar or RANS) that simply controls the residual you want to achieve.
- For unsteady solutions (DNS/LES/URANS) that makes sense only in statistical sense. For example controlling kinetic energy in time for a flow in equilibrium or to check if something is going wrong (a rapid increasing in the kinetic energy can be the start of a numerical instability).

In your case I suppose that the monitor of the time evolution can supply the evidence of the correct fundamental frequency of the vortex shedding

leflix September 2, 2012 09:50

Quote:

Originally Posted by 1/153 (Post 379775)

Once my boundary condition is right, the seatdy simulation results are good. Do I still need to monitor the convergence?

It's never necessary however it helps to check if everything is going well.If you see a residual on a given variable which increases dramatically, then you know something is going wrong with the computation of this variable. Then you can track more easily where is the bugg.
if you compute incompressible flows then I would also monitor the continuity constraint div(U)=0 to check if is satisfied....
But keep in mind too that each time you ask your program to display something on screen it delays your computation. Some people loose 1/3 of the overall computational time in writing stuffs on screen that they never check....:D

FMDenaro September 2, 2012 11:00

Quote:

Originally Posted by leflix (Post 379883)
It's never necessary however it helps to check if everything is going well.If you see a residual on a given variable which increases dramatically, then you know something is going wrong with the computation of this variable. Then you can track more easily where is the bugg.
if you compute incompressible flows then I would also monitor the continuity constraint div(U)=0 to check if is satisfied....
But keep in mind too that each time you ask your program to display something on screen it delays your computation. Some people loose 1/3 of the overall computational time in writing stuffs on screen that they never check....:D

I agree! Never, never direct output on screen, much better to decide an interval and write the residuals on a sequential file... ;)


All times are GMT -4. The time now is 10:50.