fan simulation advice
 

Hi,

I am simulating an axial fan that is forcing air through a radiator. Both are contained in a large duct. I cannot get CFX to converge steady state and have tried transient successfully but transient takes too long to solve. I have tried all the things recommended in the manual for steady state including all sorts of physical timescales. I've also tried LTS factor that improves things slightly but still doesn't get good residual values.

I have tried refining my mesh but it doesn't help. Why is there this problem with CFX and would switching to another solver help (such as Fluent)? Plotting residuals shows problem areas where the cell size is large and the flow is slow. I have even tried another fan with the same setup and it gives the same problem.

Please can someone help.

Thanks,

ijk

Unfortunately, such is the lonely life of a CFD worker. If you've tried everything else, including the things in the link Glenn posted, and your simulation still isn't converging to an acceptable level then transient is your only option. Either upgrade your computer, perform batch parallel processing (i.e., multiple computers solving the simulation), purchase more licenses so you can run in parallel or run multiple simulations at once, or connect to something like the High Performance Computing Virtual Laboratory (HPCVL).

It looks like you've already figured out your problem. Your simulation is not steady. It's an axial fan - there are probably unsteady effects from secondary flows (e.g., separation) and the like. It's not a problem with CFX; it's a limitation of CFD. Switching to Fluent won't help - they're both implicit solvers with roughly the same commercial development. If the problem areas are where the cell size is large, then refine the cell there and/or do a grid quality check (e.g., if you're using a structured grid, check the orthogonality).

Good luck.

Thank you both for your responses. My steady state residual levels are high but the quantities I am interested in are stable with steady state. Sliding mesh is much more accurate for what I am simulating anyway.

Should I expect Fluent and CFX to take the same time to solve the problem transient (for the same timestep)?

CFX and Fluent are very different technology solvers and there is often a considerable speed difference between them. In some of my recent work the difference was a factor of 30.... but it depends on the simulation so you will have to benchmark it for yourself. There is no general rule.

Glenn,

That is very interesting. I will try a few test cases. Does anyone know if Fluent or CFX would be faster for transient fan simulations?