CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   How long an analysis for steady state drag? (http://www.cfd-online.com/Forums/main/108388-how-long-analysis-steady-state-drag.html)

ABF October 22, 2012 11:35

How long an analysis for steady state drag?
 
Hello,

I'm modelling flow over a rigid body, with the aim of extracting drag and lift data. I'm curious as to whether there is a rule of thumb I can follow that will help me establish how long I ought to run the simulation for in order to be confident I'm getting the full picture?

An example analysis of flow over a steamlined body (torpedo shaped) was performed over 10s with the drag appearing to level-off after about 2-3s. However, my initial models of a sphere (as a learning, test-case) have shown the drag to appear to be level up to about 15seconds (Re about 1.5E6), before entering a complex sinusoidal pattern that I assume is due to the vortex shedding.

I'd like to be sure I'm not cutting off my analysis time too soon - I know the easiest option is simply to run the analysis for longer to see, but in the interest of processing time, this isn't a 'good' solution. Is there a rule of thumb I can follow?

Thanks
Adam

FMDenaro October 22, 2012 11:48

Quote:

Originally Posted by ABF (Post 387917)
Hello,

I'm modelling flow over a rigid body, with the aim of extracting drag and lift data. I'm curious as to whether there is a rule of thumb I can follow that will help me establish how long I ought to run the simulation for in order to be confident I'm getting the full picture?

An example analysis of flow over a steamlined body (torpedo shaped) was performed over 10s with the drag appearing to level-off after about 2-3s. However, my initial models of a sphere (as a learning, test-case) have shown the drag to appear to be level up to about 15seconds (Re about 1.5E6), before entering a complex sinusoidal pattern that I assume is due to the vortex shedding.

I'd like to be sure I'm not cutting off my analysis time too soon - I know the easiest option is simply to run the analysis for longer to see, but in the interest of processing time, this isn't a 'good' solution. Is there a rule of thumb I can follow?

Thanks
Adam


Hello, to give some answer you should address the type of formulation you are using ... Are you using URANS or LES?
As general rule, you can control the total kinetic energy. When it is statistically steady (that is after the numerical transient is finished) you can take the samples to get a statistical average of the fields until your statistics are steady.


All times are GMT -4. The time now is 18:44.