Extreme drop in drag after 70 iterations
 

Hi,

I am running a transient simulating of drag over a body at supersonic speeds. I have added a monitor point to monitor the progress of drag throughout each iteration. The drag seems to settle around 0.7, which is what is expected. However, after a few more iterations, the drag drops to below -1000. Anyone have any idea where this error could come from?

Thanks!

Did you create monitoring points around your object, where you monitor p,v,T,Mach etc? From that you might get an idea what is really happening during your calculation.

Also, you could create a backup file at iteration 70 (drag 0.7) and at iteration 80 (drag -1000) and compare the results in Post.

This is almost certainly the solver starting to diverge. Pretty soon I would expect the simulation will totally diverge and crash with a floating point error. See the FAQ: https://www.cfd-online.com/Wiki/Ansy...do_about_it.3F

Thank you both for your answers! I will do some modifications and try again.

I was also wondering about the size of the fluid domain. Is there any guideline concerning the ratio between the size of the object and the surrounding fluid domain? I haven't been able to find any definite answer.

The proximity depends on the flow regime so it is hard to give guidelines. But as your object is supersonic that means it is likely to have a very long wake and shockwaves. This means the downstream boundary will need to be a long way downstream and the side boundaries quite a way back too. For subsonic models the downstream boundary 20 times the object size down stream and the side boundaries 5 times the object size away is a starting point, but it is quite likely that will be too close for your case.

The best way to determine it is to do a sensitivity study. Generate a mesh with the downstream boundary 20, 40 and 80 times long and see what the difference between the results is. This will allow you to choose a length to give the accuracy you require.

How about if you slice the object and domain along a symmetry plane and set the side boundary condition as symmetry in order to save computational time? Would you expect that to generate different results than including the whole geometry and fluid domain?

Again, that is problem dependant. If the flow is symmetrical this is a good idea any you should do it. If the flow is asymmetric then you should not do it. Note that symmetric geometries can generate asymmetric flows, such as the Von Karman vortex street (https://en.wikipedia.org/wiki/K%C3%A..._vortex_street)