Dear All,

I have a (HVAC type) flow driven by fan heaters and buoyancy. The fans heaters are in the top half of the domain. This cases the top half of the domain to be hot, with velocities ~2 m/s. The bottom half of my domain is cooler - due to thermal stratification, and the velocity is almost zero. The top half of the domain converges OK, but the residuals in the bottom half stay high.

1) Any tips of how to converge this type of simulation? How can I encourage the bottom half to converge quicker?

2) I have been using 'Local timescale' control to try and improve convergence. But the CFX documentation does not say how the timescales are computed at different points in the domain.

- Does anyone know how the local timescales are computed?

- Does anyone know what the 'Local timescale factor' really means? What values do you use? (CFX help is cryptic about this too!)

- Can I specify a maximum value for the local timescale?

The reason I ask is that I want to use a large local timescale to help convergence, but not so large that buoyancy causes instability, etc. At the moment it is causing me problems. This local timescale factor seems a bit mysterious!

Many thanks.

Best regards, Andy

Use automatic timescale and make a transient simulation varying the total time to check the convergence.

Hi Andy,

The local timescale gizmo should be used to approach convergence but not as the final run to convergence. You should still use some physical timescale component to do the final run to convergence.

I would use the largest physical timescale which is still stable. In simulations like this you often get large scale transient instabilities at low frequencies (of the order of a few seconds) so you need a timescale several times that to have a hope of convergence. Even then they often don't converge nicely and require a transient approach to resolve the instability to solve.

Glenn Horrocks

Glenn, Roger,

Thank you both for your suggestions. I will investigate them.

A transient approach has helped convince me there is no major mesh problems, etc.

I think it is likely that I'm seeing low frequency instabilities as suggested. I will now try upping the timescale to try and avoid them.

Regards, and thanks again. Andy