First, a simulation using powder elements should be run to which to compare the convection approximations.
From a command line run:
$ pan -b pdr
The analysis progress is written to file pdr.out.
After the analysis completes, make note of the CPU wall time from the log file. This case takes 1 minute, 23 seconds to complete. Actual CPU times will differ.
Now run the global convection model, which approximates losses into the powder and ambient environment as a uniform heat flux of 25 W/mm2 K.
$ pan -b global
The log file returns a CPU time of 24 seconds for this simulation. CPU times may vary.
Next run the first advanced convection approximation model which applies regional convection values. These values are as follows:
Powder Bed Substrate Convection – *PBSB = 150 W/mm2 K. This approximates losses from the build plate base into the build elevator. A high convection boundary is necessary to adequately model conduction heat losses as a heat flux.
Powder Bed Substrate Sides Convection – *PCSS = 125 W/mm2 K. This approximates losses from the build plate sides into the powder and walls of the powder bed machine. This is also an applied heat flux simulating conduction losses, but as there is a thin layer of powder between the solid build plate and the solid machine walls, the rate of heat transfer is less than for the build plate-build elevator surface.
$ pan -b regional
The regional log file shows a CPU time of 23 seconds. CPU times may vary.
Finally, run the second advanced convection approximation model, which uses the same values as the regional case, but has an additional control card, *TCNV. This card assigns different convection values, based upon the thickness of the component. These values override the values specified by *CONV. For this example the thick sections will be given a flux of 5 W/mm2 K while the thin sections, which lose heat more rapidly, will have a heat flux of 20 W/mm2 K.
$ pan -b thickness
The thickness simulation takes 25 seconds to complete.