Clarification regarding Intelligence Convergence Control

[OmkarJ]

Hi, 

I was wondering how ICS works exactly. I understand that it runs as a pseudo transient solver and adjusts the timescales to prevent the divergence.

But,

1. does it use the same timescale throughout the domain?
2. or different timescale locally depending upon the local velocity/cell size  etc? 

#1 would mean that globally, the timescale would be influenced by largest velocities/smallest grid size in the domain. This would again mean that for parts of the domain where local velocities are relatively higher, the smaller timescale (globally) chosen will result in slow propogation of result. On the whole, this approach would prove to be drastically slow for large no. of cells.

#2 would be beneficial in the sense, local timescale would mean solution propogates at different rates throughout the domains, with high velocities opting for smalelr timescales and small velocities opting for larger timescales. Thus, the divergence is avoided, at the same time the solution propgates adequately. 

Either way, would it be beneficial to turn ICS off even in steady state, after initial turbulent and numerical instabilities have calmed down?

Lastly, how does ICS affect the under relaxation factors chosen for p, v, T etc?

Would like some views on this discussion.

Thanks

OJ