One thing you can do to null out the sound is to track everyone on the
    factory floor within earshot of the machine.  Once you know their
    location to within a few inches (particularily the location of each
    head, or even each ear) then you can solve a series of simultaneous
    equations which will determine the phase of sound you should produce
    from a speaker near the machine.  This equation will give the best noise
    dampening for each person, given that a few aren't in optimal positions.

Is this possible?  Ten or twenty years ago I would have labeled such a
suggestion as a obvious joke, but with todays gigaflop DSPs and what
not, it may have entered the realm of possibility.

It's a good idea in general to re-examine the set of dismissed
"impossible tasks" every couple of years to see if they've become
suddenly become possible due to technology advances.  I can see this all
the time in code at work - some bit of ugliness that was difficult to
make better, simply left that way, even after the infrastructure to do
away with the "ugly" part has been in place for a while.


BillW