A nice idle computer begs for some numbers to crunch, so here's some recycled
results from about a year ago.  To really do 'good' CFD would require a
complete overhaul, but it's perhaps marginally better than the semi-Lagrangian
technique.  I think the channel height Reynolds number is about 250, which puts
it beyond the realm of accuracy.  FYI, it turns out a general
speed*length/viscosity (= Reynolds #) fully defines the problem, so if you've
solved one Re, you've solved for all whole class of flows.  Further FYI, the
sizing in this problem is for air at about 1m/s and a cube of 2.5 mm!  That's
why turbulent flow is such a challenging problem.  Think of analyzing a 747...
(I think an estimate was around 760 Million years for a full simulation.
Ouch.)

 - Ricky

Post a reply to this message

Attachments:
Download 'output.mpg' (685 KB)