Let's venture into some of the
more esoteric robotic drives, starting with
the strain wave gear or known by its commercial name
harmonic drive, invented in 1957 by C.W. Musser.
It is a nifty little gadget but it's really
hard to wrap your brain around it. While this
series of courses have always been geared towards to
the home hobbyist, I recognize that there are
students enrolled who want to get into robotics
professionally, and so the next few lessons on these
esoteric drive mechanisms are surprisingly important. ASEA
used the harmonic drives extensively in their
first robots, and as you'll see, with good reason.
Harmonic drives are also used in aerospace, for
example the drive train of the Apollo moon rover used
electric motors geared down through harmonic
drives to drive the wheels. The Skylab, which was
the first space station put in orbit by the
United States, it used harmonic drives to deploy its
solar panels. My college instructors had a
working, see-through plastic harmonic drive for
demonstration. I sat there and played with that thing
extensively, trying to comprehend it and I'll
be honest with you, I never did figure it out.
It wasn't until many, many years later that I
understood the principle of the mechanism. We
didn't go into understanding how this drive
worked in our college class, it didn't really matter.
Either it worked or it was broken and we replaced
it. But these things were so incredibly reliable that
no one in the college or in industry could
recall one ever failing. But obviously they are
important for those of you who wish to enter into
robotics on a professional level, because you
will encounter these in industry. So here you can see I have two
racks a red rack which will serve as the stator
and a blue rack which will serve as the output.
The blue rack has two teeth fewer than the stator
rack. You can see that as a result, the teeth only
line up in two locations along the length of
the racks. I have a third rack here which
is flexible and is called the flexspline. The flexspline rack can now only
mesh together with the other two racks in two
places where the teeth are aligned. If I force the flexspline into
the other two racks here where the teeth are not
aligned, it acts as a wedge, forcing the teeth to
align with each other, moving the blue output rack. The
red stator rack would be mounted to say, the
robot body, or a foundation of some sort. The
output rack is mounted to the load, and by forcing the
flexspline into the two racks, the output rack will
move to align its teeth with the stator and
flexspline teeth. You remember how powerful the
simple machine the wedge was, right? Well look at
this the flexspline is acting as a wedge
to force the output rack to move relative to the
stator rack. So now if I take a bearing and
apply a force against the back of the
flexspline, and move that bearing along the length of the
flexspline, it uses the teeth on the flexspline to
work as wedges all along the racks, forcing the
output rack to move. So my input force only has to
move this bearing laterally, using the huge
mechanical advantage of the wedge to move one rack
relative to the other. Of course, it is reversible as
well, by merely changing the direction of the
bearing movement. This setup actually has a
mechanical advantage of about 50:7. The input moves
about 50 millimeters, causing the output rack to move about 7 millimeters. There are, of course, TWO places
where all the teeth line up. So if I had two
bearings moving in unison, it would double the
effort by making use of two wedges forcing the racks to
move at the same time. Notice as well, that if the
bearing is stopped, the wedge is still in place and
there is NO movement allowed between the stator and
output racks. It has zero backlash! Wow this is
shaping up to be an awesome drive mechanism, how can
we modify this to make it even more useful? If I take these and bend them
all into a circle... and I put a cam of some kind as
a rotary input which simply forces the cam or
bearings out in opposite directions, this will
drive the flexspine in a rotary fashion with the
input. The cam will force the teeth of the
flexspline into the teeth of the stator and output ring
gears, and as it turns, it will cause the stator and
output ring gears to turn relative to each other. It is this flexspline that is
the most complicated part of the harmonic drive, but
also the reason for its huge success. It is
difficult to make one of these in your home shop, but it
can be done. 3D printed harmonic drives have
been made, but finding a hard yet flexible plastic for
the flexspline is difficult. You can probably make
one, but it would be unreliable at best. I have here a demonstration unit
that was 3D printed from a design provided
by Simon Merrett on YouTube, he was inspired by a
post on Hackaday where one hardware hacker came
up with the idea of using a timing belt as the
flexspline in a 3D printed harmonic drive. Here you
can see the timing belt, it's been flipped inside
out and it's locked into the green stator gear, and you can see the bearings
which flex the timing belt against the orange output
gear. So this bolt going through the middle goes through ball bearings on the stator and output gears
so they both freewheel on the bolt, but the
input rotor is double lock-nutted onto the bolt
in the middle. So as you turn the bolt, it turns
the input rotor, which moves the bearings around
the inside circumference, which moves the
wedge into the teeth of the orange output gear. The
orange output gear has two teeth fewer than the
green stator gear. Let's add some input power so
you can see it in action. This drive has about a 45:1
reduction ratio. Notice we can also reverse the
direction! There's also zero backlash there
between the two gears like, none. Let's take a look at an actual
harmonic drive from my Fanuc robot. Here you can see
the flexspline. It's got very fine teeth, and yes, it
is flexible but not much! Literally the flexing that takes
place when in operation is on the order of a
millimeter or two. In this particular harmonic
drive, they've only got the stator gear. There is no
output gear. The output comes off the flexspline.
So the flexspline has a few teeth fewer than the
stator gear, and a cam is rotated by the
drive motor which forces the flexspline teeth out
to engage with the stator teeth. Because of the
difference in tooth count, the flexspline rotates at
a speed considerably slower than the
rotation of the cam. So while very difficult, you CAN
make strain wave gear drives at home. For those
of you in industry, they're actually a really
impressive mechanism and there's really no maintenance
involved they're either broken completely, or
working. They do wear out over time and build up some
backlash. However, most factories replace their
robots every couple of years sometimes as often as
every year, just because the wear and tear on the
robots causes them to become more and more
inaccurate over time. And accuracy on an assembly line is
of paramount importance. So you'll see most
factories opt for a preventative maintenance
attitude and simply replace the entire robot
assembly line.
