Dear Tay Wu
I fully agree that nothing beats measuring in a controlled environment but if You have to try to compensate for thermal expansions -
what alternatives are there?
The approach I described is done for Aluminum
and yes, there are three invar bars. The advantage is that effects of, for example bending, is determined, constantly, by measuring what happens with the machine, instead of trying to measure just temperature and then trying to calculate what the effects would be.
So, a temperature controlled room is best but I still think the method described is interesting!

Regards Karl-Johan Karlsson