Hi everybody! I am writing a code for 3D unsteady,turbulent,compressible flow in cylindrical coordinates. The problem that I am facing for many months (!!!) is that at thita=0 and 360 degrees the results that I am having are not the same (or not almost the same), when the variables are small. Actually they are very different. The bigest problem is with the pressure correction, because the values that it takes are quite small (10E-3). Beacuse of this fact the velocity correction at 0 and 360 degrees is very different and also the d(P)/d(theta) is very different (for example: at 0 degrees: 0.001 and at 360 degrees: -0.0005). These values are when the code has converged. During the early iterations the difference is very large (5 and -12). The way that I am implemending this, is that I use ghost cells at 0 (I=1) and 360 (I=NI) degrees that take the values of I=NI-1 and I=2 respectively. These ghost cells are actually the boundary cells and have exactly tha same characteristics (volume, dimensions etc.) Do you find this way correct,or is there a better one? Has anybody faced the same problem with cylindrical or spherical coordinates? I have also tried periodic or cyclic TDMA but the results were exactly the same. Any thoughts? Thank you in advance!

You might try to have one more ghost cells at theta=360 and go for I=0 (theta=0), NI+1 (theta=360+2*dtheta) and then implement the conditions as

NcellI=NI+2

phi(k,j,NcellI)=0.5*(phi(k,j,2)+phi(k,j,NcellI-1))

I am using 3-D FVM in cartesian system as a 2-D solver and its working for me for 2-D flow over cylinder.

Now I have one ghost cell at 0 degrees and one at 360 degrees. You suggest me to have two ghost cells at 0 degrees and two at 360 degrees? The solver (TDMA) for the cylindrical coordinates in comparison to the solver for cartesian coordinates will change or will be the same? The flow that I am simulating is inside a cylinder and not around a cylinder. Waiting for your answer! Thank you

I solved for external flow with the two ghost cells.... you might try it too internal also... I guess there wont be any change in the solver but only implemenetation of the condition will change....

Best regards.... ramp

I have sent you my solver to see if it is the same with yours. If not, can you send me the solver with the changes for the cyclic conditions and its implementation that you have? Thank you

Sorry, I forgot. My mail is gkosmadakis@gmail.com

hi george i work on the problem of laminar mixed convection in the entrance region of a semicircular duct. so my problem is three dimenstinal and i work with SIMPLER AND FVM

i do not understand your equation and i will asked to my supervisor. can i ask you george in which is your university?

because your problem is exating and i really want work with turbelent and compressible flows.

really if tell me i will try to join your reaserch.

yours yousef