Some people (myself included) like to approach this question using the “chipload” value, that’s the thickness of the chips you are getting when cutting with a given combination of RPM, feedrate, and endmill.
The idea is that for a given material and tool, there is a sweet spot for the chipload value that can be reached using a variety of RPM/Feedrate/endmills combination.
Chipload in inches = (Feedrate in inches per min) / (Number of flutes x RPM)
In your case, the chipload value that seems to grant good results is:
chipload = 20ipm / (2flutes x 10.000RPM) = 0.001"
(which does not surprise me as 0.001" is kind of the magical chipload value for cutting aluminium, and more)
The chipload value does not depend on the endmill diameter, so no need to compensate for that (however, the MAXIMUM chipload that a given endmill can take does depend on its diameter). In your case, if a 3.175mm endmill could take it, a 3mm endmill will take it just fine.
So since you are going from 2 flute to 3 flute, the only thing to do to keep the same sweet chipload of 0.001", is to adjust the feedrate:
new_feedrate = chipload x RPM x nb_flutes = 0.001" x 10.000 x 3 = 30ipm
Don’t forget the RPM in the formula though, it’s just as important to adjust your feeds and speeds. Sooner or later you will find a case where you would have needed to increase RPM, but are already maxed out
The only caveat is that this presumes that the flute geometry and size will support clearing the chips. That is going to depend on the actual cut geometry (all other things being equal). A strategy where the tool is fully engaged in a slot will be different than a surfacing cut, as the chips will have somewhere to go when surfacing, but are more likely to get packed up or recut in a slot, regardless of the axial engagement for the pass.