For steel the coefficient of linear expansion is about 12 x 10e-6 per
degree C. Figure the max temp difference you can achieve while
assembling the worst case fit (oversized shaft tolerance with undersized
hub tolerance. If you put a 1" shaft in some liquid nitogen at -196C and
heat the hub to the max your home oven will go (say 270C) you will get
..0055" of total expansion/shrinkage. The max interferance fit you can
get in practice has a lot to do with how accurate your machining is plus
how well you can assemble the parts before they start normalizing their
temps.

We did one aluminum prototype in two pieces that was supposed to have
..002" clearance with the temps we were using. We wound up using a press
to assemble since the aluminum conducted the heat so fast. But the parts
were bonded together1

Cheers.


Terry wrote:
I like the idea of shrinking the flange onto the tube. I can easily
machine the flange I.D. to be a couple of thousands smaller than the
tube O.D. Then I could use a hotplate to heat up the flange then slip
it into place and let cool. I could then put some very small tack
welds in place.

How much undersized would you recommend making the flange I.D. compared
to the tube O.D. for a permanent fit with no chance of slippage?

If I can't true up my existing assembly, I will go this route on the
next one. Thanks.