CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Initial drag for cylinder in cross-flow (http://www.cfd-online.com/Forums/main/102562-initial-drag-cylinder-cross-flow.html)

orxan.shibli May 28, 2012 16:26

Initial drag for cylinder in cross-flow
 
Hi everyone!

I am simulating flow over cylinder at Re=100 with my own code. Setting time step to dt=1s, I monitor initial values of drag coef. and it solves it without problem. But when I set it to say dt=0.001 then initially drag coef. rises to very high values (hundreds) and then it becomes negative even. It is not important whether it will be better as time passes since I am also trying to understand why decreasing time step causes this kind of problem.

Other specifications:
  • Unsteady
  • Fully Implicit, 1st order Euler temporal discre.
  • 2nd order upwind spatial discre.
  • Max. Inner Iterations = 10
  • 2D, laminar
  • Initially velocity and pressure field is zero

Any idea?

cdegroot May 28, 2012 20:46

You are starting the flow from a standstill, so that takes a big pressure drop to accomplish. That is one reason for the high drag. Also, you are only doing 10 inner iterations, which probably means it is not converging within each time step. In that case, the initial results don't mean much anyways.

orxan.shibli May 29, 2012 06:44

Dear Chris

OK but this should be true for both time steps. That is why I decreased time step however it became worse unexpectedly. Also I tried to make inner iterations 100 but that only caused to reach high values faster.

cdegroot May 29, 2012 11:06

Well, your small timestep is 1000 times smaller than your large time step, so I would guess that the large time step just "steps over" a lot of the initial transient stuff.

Also, imagine in reality having completely still fluid and then instantly accelerating it to whatever value you have specified at the inlet. Not going to happen. It will take some (maybe small) amount of time to accelerate. In your simulation that time to accelerate is your first time step. Thus, I believe a smaller timestep will imply a larger initial pressure drop and also drag forces. In any case, these initial transients will die out eventually.

orxan.shibli June 3, 2012 18:43

I found the problem. It was due to time-step dependency of original Rhie and Chow MIM. Hence, it is better to use modified version:

"Discussion on Momentum Interpolation Method for Collocated Grids of Incompressible Flow", Yu et al.

cdegroot June 4, 2012 02:10

Quote:

Originally Posted by orxan.shibli (Post 364531)
I found the problem. It was due to time-step dependency of original Rhie and Chow MIM. Hence, it is better to use modified version:

"Discussion on Momentum Interpolation Method for Collocated Grids of Incompressible Flow", Yu et al.

Very interesting. Thanks for sharing.


All times are GMT -4. The time now is 00:20.