Using CFX5.7 for transient mod....How to define timestep??If it is the physical time that the flow passed the domain????

Your timestep needs to be small enough to resolve the transient features of your flow. Generally this means that you should converge within 3 to 5 coefficient loops within your timestep.

A timestep equal to the resident time would be far too big.

Regards, Robin

Under a laminar flow simulation, what be the possible reason that would cause the solution to diverge with smaller timestep?

Good guidance would be to start with the courant limit. This is the smallest element length divided by the expected velocity through that element (i.e. the time it takes a fluid particle to pass through one mesh element). For an accurate time marching solution, you certainly don't want anything larger than that.

Highly fluctuating oscillations may require an even smaller step size, but it's a good place to start.

Jeff

Thank you for your suggestions,,,

Hi,

To answer Chien's question:

Smaller timesteps can lead to divergence as you can now resolve more flow features. For modelling a bluff body with a large timestep can just show a large wake, whereas the same model with a fine timestep can resolve the details of the oscillating vorticies in the wake. Resolving these vorticies is a tougher ask numerically than the bulk flow field, so if your simulation is not very stable it could cause divergence.

Glenn Horrocks