CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   CFX solutions diverges (

b.shuvayan January 14, 2013 05:37

CFX solutions diverges
HI there!

I am trying to solve a problem in ANSYS CFX for the last 2 months but it just dosen't get converged. The problem statement is as below.

I need to model flow around a body ( a 3D triangle). The body itself is inside another body ( a cube of dimension 600 mm each). The triangle is kept at the centre. One side of the cube is inlet and the other is outlet. I need to study the flow variation around the body (i.e triangle). Its all supersonic flow with Mach no. close to 8.

I need to create structured mesh which I have done with the help of slices. The problem is the solution does not get converged at solver manager and shows error 255. I have tried for almost 2 months. Need some urgent help.

Thanks in advance

MSHowes January 14, 2013 15:08

Re: CFX soutions diverges
Hi Shuvayan,

You haven't said enough for anyone to begin to guess what's happening. Are you running in transient or static mode? How many time steps or iterations into the simulation does the run crash? Are you getting any kind of error message regarding the linear solver or a bounds error? Have you tried putting out frequent .bak or .trn files so you can see what's happening?

b.shuvayan January 15, 2013 11:41

Okay let me explain.

The problem statement is in static mode. NO of iterations is 1000 and auto timescale has been selected. I receive an error message 255 towads the end and it just hangs up and closes. NO of iterations it carries on is max 100 and crashes. I have looked at .bak file but the results are not at all satisfactory. The velocity , temperature and pressure contours are all not proper.

Can you suggest anything. ??:confused:

MSHowes January 15, 2013 12:20

CFX Solution Diverges
I'd take a different approach to this problem: your initial conditions are probably not close enough to the correct solution for this to work in static mode. I'd run it as a transient case and I'd ramp up the velocity by defining the input velocity as a CEL expression that is a function of time. You can use the "min" function in the CEL expression to prevent the velocity from continuing to increase once it reaches your target value.

Start with a small timestep and use adaptive time-stepping; you might need to play around with the initial timestep and the maximum coefficient loops you're allowing to see what will work.

Just my opinion; I've found transient mode to be successful when static runs have failed. The solution will eventually converge to the steady-state (static) solution if it works.

By the way, when you say you have a 3-D triangle, do you mean a prism? A prism has a triangular base and rectangular sides. Just curious.


singer1812 January 15, 2013 12:39

You can also try ramping up the inlet to your set point even within a steady state run.

Mach 8 is getting pretty high. Are you sure you arent getting dissociation or chemestry effects? Are you sure CFX is a good code to do this in?

ghorrocks January 15, 2013 21:50

Another comment: For steady state runs local timescale factor can be very useful for high mach number runs. use a factor of 5 or so.

But Edmund's point is important - most people consider Mach 5 as the limit of when dissociation effects become important. So you probably have significant dissociation and that means CFX cannot do it.

b.shuvayan January 17, 2013 09:59

Well that's what i have been told about cfx by my prof. Fluent won't be of much help I guess. The problem lies perhaps in meshing because I have solved similar set of problem in CFX with Mach No. 8, only the geometry was little different.

ghorrocks January 17, 2013 21:41

At these sort of Mach numbers even a tiny little pimple will generate a shock wave which travels out, reflect off stuff and bounce around. This can cause a big shock wave to move, and suddenly the fluid flow is totally different. Small differences can be very important.

b.shuvayan January 18, 2013 08:08

which is why I guess it hasn't solved since 2 months. :)

What do you think the no. of nodes or elements should be for a cube of 600 mm side?

All times are GMT -4. The time now is 09:59.