Hi All,

I am running a duct type of problem in CFX5.4.1. I was able to get the problem to converge using the auto time step of 5E-3. I am now trying to run the problem in reverse as well, switching the inlet and outlet and changing the initial guess accordingly, but the auto time step now is 5E+15, which is a bit troublesome. I have tried lowering it back to 5E-3 and running, but it keeps giving me a fatal overflow before the first iteration. Does anyone see anything obvious settings that I may have missed that would give me such a large auto time step. (Also, what are the units on the auto timestep reported in the .out file?) Thanks,

Carrie

I think that you have probably made a mistake in your problem setup. Check it carefully. Especially check that you changed the initial guess option from "Automatic" or "Automatic with Value" to "Value". If you don't do this then CFX-5.4.1 picks up your old values as an initial guess, the flow is going the completely wrong way, and the solver probably blows.

Also, if you are just running a simple pipe I would suggest that you use Physical Timescale rather than the autocomputed timescale. Set the value to something on the order of 1/3 -> 1/5 of a characteristic length/velocity scale for the problem.

Neale.

Hi Carrie,

Sounds like you model dimensions are off. The default solver dimensions are SI, but you can check by looking at the units reported in the CCL echoed at the top of your .out file. The wrong dimensions will also cause the fatal overflow.

Robin