Off and on for the past couple of
years, I've been working with nylon monofilament artificial muscles,
trying to reproduce the results seen by professional researchers at a
hobbyist level. My experience with the topic is strewn across a
number of articles on this blog. Since there's been a fair bit of
interest, I thought I would roll up everything I've learned so far
into a comprehensive guide.
Introduction: what are these things?
In general, an artificial muscle is an
actuator which produces motion by contracting or expanding along its
length, or by rotating. Artificial muscles made of nylon
monofilament were revealed in early 2014, when an inter-university
team published the report “Artificial Muscles from Fishing Line and
Sewing Thread” in the journal Science. As suggested by the
title, nylon monofilaments are easily obtained by the layperson in
the form of fishing line and nylon-core conductive thread. When the
nylon is coiled in a specific manner and heat-treated so that the
twists are made permanent, it becomes a muscle that responds to
changes in temperature by (reversibly) changing its shape.
Activation of the muscles can be achieved in a variety of ways.
Blowing on them with a hair dryer or heat gun, immersing them
alternately in hot and cold liquids, and putting them in contact with
an electrically heated wire are all methods that have been used
successfully. For an average or general application, I favor hot
wire muscles as the least bulky, messy, and complicated of the
possible options, so I'll be focusing on them rather heavily in the
discussion that follows.
What nylon artificial muscles are
good for
Reading some of the articles about
these devices could make you believe they're a miracle, but you
shouldn't necessarily expect them to be the perfect solution for your
robotics project. When compared to more traditional actuators like
servos, they have distinct attributes that could make them an
excellent choice for some applications, and a complete flop for
others. Many of these would be common to thermosensitive artificial
muscles of any type; a few distinguish nylon muscles from their chief
competitors, shape-memory alloys such as nitinol wire.
Advantages:
*Light. Their power-to-weight ratio is
very good – exceeding that of human muscle tissue.
*Silent.
*Inexpensive. The cost of a single
nylon muscle could probably be measured in pennies. (Advantage vs.
nitinol wire)
*Easy to manufacture at home, and thus
fully customizable to your application's needs. They can be made in
a variety of form factors that trade actuation force against maximum
actuation distance. (Advantage vs. nitinol wire)
Disadvantages:
*Slow. The muscles I've built take
minutes to go through a full actuation cycle. Much faster cycle
times have been achieved by others, but it seems that either
immersion in water or active cooling (by fans, for instance) is
required for best results. The thicker the nylon, the slower the
heating and cooling process will be.
*Inefficient. If you want good
performance during the heating cycle, these muscles are rather
power-hungry.
*Environmentally sensitive. Since they
are actuated by heat, nylon muscles will respond differently
depending on the ambient temperature, requiring calibration and/or
feedback if precision movement is desired.
*Delicate and unreliable. It doesn't
take a very high temperature to set the nylon into a new shape, so
care must be taken when actuating the muscle to avoid its accidental
destruction. Even high ambient temperatures (as in a closed house on
a summer day) can sometimes do bad things to them.
Muscle materials
The research team that published the
Science article used a type of filament called nylon 6/6.
Supposedly this nylon grade is often used for fishing lines, but the
packaging isn't going to be that specific – if you want to
guarantee that you have real nylon 6/6, there's a good chance you
will have to mail-order it, at additional trouble and expense. I
opted instead to just buy some things and try them, and so far I have
yet to find a type of nylon that doesn't work. I've used a couple of
different fishing line brands, beading string, and even hedge-trimmer
cutting line and gotten them all to actuate. One hitch to beware of
is that not all fishing lines are made of nylon – the fluorocarbon
type is also popular, and I've yet to hear of someone making a
working muscle out of that.
For your first experiments, I recommend
a filament that is fairly thick, but not too thick. A 711 um
diameter fishing line worked well for me, and continues to be my
favorite thing to use. In my experience, thinner nylon is less
forgiving – it's more likely to infuriate you by snapping during
the coiling process if the tension you put on it isn't exactly right,
and it's easier to overheat and destroy it when you start playing
with your new muscles. However, I don't recommend the really
thick stuff (deep sea fishing line, trimmer line, or 3D printer
filament) for your first try either. 2 mm filament calls for a heavy
weight to tense it during coiling, and if it should happen to snap or
come loose, watch out for your eyeballs. It soaks up a lot of heat
and is harder to get a response out of with the application of
reasonable amounts of energy. So get a “happy medium” filament,
and try the extremes when you're more experienced.
If you're planning on including an
electric heating element in your muscle, you must also choose an
appropriate wire. The best wires have a resistance in a range that
will give you about as much current as your power supply can safely
offer. They also have to be as flimsy/flexible as possible, as any
mechanical stiffness in the wire will present resistance to the
motion of your muscle.
I've done much of my muscle work at 5-6
V, with up to 1.2 A of current available. At this point in the
parameter space, I've had the best success with very thin copper
wires. Magnet wire – the copper filament you can get out of old
motors and transformers – heats up well, but is unacceptably
fragile. Nichrome (30 AWG) is on the opposite end of the spectrum:
strong for its diameter, but not as good at quickly heating up and
cooling off, and possibly stiff enough to impede its muscle. Wires
made out of many individually insulated copper strands, such as
tinsel wire and litz wire, are the best – reasonably resilient,
with good heating and cooling performance. Unfortunately, they are
hard to purchase in small quantities. Someone was distributing litz
wire on Ebay at the time I went shopping, but the only tinsel wire I
could find is the small amount I scavenged from a broken pair of
headphones.
Another possibility that I haven't
personally tried is coating the nylon line in conductive paint. I
was wary of this option because silver-based conductive paints run
expensive, and the carbon-based paints/inks have a high resistivity
as compared to thin copper wires, so I was concerned they wouldn't
carry enough current for good heating performance at the voltages I
was working with. A different problem with this approach was
reported to me by @inventor42 on Twitter: the coating of paint is
liable to crack, and a girdling crack anywhere along the length of
the muscle will create an open circuit. The technique of using a
silver-based coating for heating was successfully applied by the
authors of a 2014 MIT paper. The paper notes that the coating must
be even for good results.
Perhaps the simplest of all options is
the nylon-core sewing thread. It already has a heating element built
in: just run a current through the conductive layer. Downsides
include a limited range of thicknesses and a tendency to be
expensive. I haven't used it myself, but it comes recommended by the
original research.
Muscle types and terminology
The coiling process for any nylon
muscle proceeds in two stages. During Stage 1, the nylon is pinned
at one end, and the other end is rotated, twisting the filament about
its own center. Stage 2 forms the twisted nylon into loops. There
are two ways of doing this. First, if one keeps the nylon under
tension and continues twisting it after the manner of Stage 1,
eventually it will begin to form itself into a series of densely
packed, telephone-cord-like coils. I refer to muscles made in this
way as “self-coiled muscles.” Second, one may produce the loops
manually by winding the twisted nylon around a rod. I refer to the
products of this technique as “rod-coiled muscles.” The loops of
self-coiled muscles have a predefined inner diameter that depends on
the thickness of the nylon, while the loops of rod-coiled muscles
have an inner diameter that is set by the size of the rod.
The chirality of a muscle is also
important. Muscles whose Stage 1 twist and Stage 2 loops are formed
in the same direction are homochiral; they contract (shorten) when
heated. Muscles whose Stage 1 twist and Stage 2 loops are formed in
opposite directions are heterochiral; they expand (lengthen) when
heated. Self-coiled muscles are always homochiral; rod-coiled
muscles may be of either type, depending on which direction the maker
chooses to wind the nylon around the rod.
The muscle manufactory
Start by cutting your chosen nylon to
length. You can estimate how much nylon you need for a muscle with a
given number of coils using the following formula:
ℓ = π(drod
+ dnylon)Ncoils
Multiply ℓ by a factor greater than 1 to account for the amount of length lost during Stage 1 twisting. When I measured the Stage 1 length reduction for the 711 um Trilene Big Game fishing line, that factor was 1.154, but you may want to experiment with your own materials. The actual resting length of the
finished muscle depends on the spacing between the coils, which you
can adjust during heat treatment. If you are planning a self-coiled
muscle, replace drod with the inner diameter of the
naturally forming secondary coils. In my experience, this is always
less than the diameter of the nylon itself; to get an exact value,
you may need to make your first self-coiled muscle a test case to see
how your particular nylon behaves. Make sure to cut at least 6 cm of
extra nylon for “tails” at the ends of your muscle (these will
not be part of its working length).
Next, you will probably want anchors
for the ends of your muscle. These will be going into the oven with
the muscle later, so whatever you choose, make it something that
won't burn or melt. I use uninsulated paper clips. Tie the nylon to
each anchor with a sound knot and tighten it with pliers, leaving a
good length of “tail” beyond the knot – the nylon will be under
a lot of tension later, and it loves to slip.
If you plan on including a heating
wire, now is the time to add it. Cut your wire, tin the ends, slip
each end through one of your anchor knots, and tie it around one of
the loops of the knot. When you are finished, the wire should lie
parallel to the nylon, and the length between the knots should be the
same for nylon and wire.
Some hobbyists complain about having
difficulty soldering litz or tinsel wire. Since each strand is
individually coated with enamel, the insulation can't be stripped,
and must be melted off in order to tin the wire. I cover my work
surface with an insulating pad and some glossy scrap paper that I
don't mind scorching, and I press the end of the wire between the
soldering tip and the paper for a few seconds. Then, maintaining
pressure, I scrape the tip of the iron toward the end of the wire.
If you're doing this right, you might see a visible streak of goo
transferred to the paper (that's the enamel coming off) and the end
of the wire should turn copper-colored and shiny. After a successful
scrape, flip the wire over and repeat on the opposite side. Then
coat the wire end in solder.
Now you are ready to twist.
The nylon must be put under tension in order to do this successfully.
If you don't stretch the nylon hard enough during the coiling
process, it will tend to form “whiskers” – by which I mean it
will start bunching up into strange-looking coils that extend
outward, perpendicular to the line, and become a useless mess. If
you stretch it too hard, it will snap partway through coiling. The
recommended coiling tension is 17 MPa (obtained from the original
research paper). The thinner your nylon, the more tightly you need
to adhere to this prescribed tension to avoid whiskering or breakage.
Don't be fooled by the nylon's pound test rating (if it has one);
during the strain of the coiling process, it will break while holding
far less weight than that!
The easiest way I know of to
get the right tension on the line is to suspend a known mass from it.
You can compute the mass you need from the following formula,
plugging in 17 MPa for T:
m = (π(rnylon)2T)/g
Hang
the tensioning mass from that convenient paper clip you put on the
lower end of the muscle.
You
could simply hang the upper end from a fixed point and twist the
muscle manually, but that would be very tedious. I strongly
recommend some sort of motor to spin the upper end of the muscle,
while the suspended weight is constrained with a horizontal stop so
it cannot rotate. A power drill, fixed to a surface well above the
floor, is ideal for the purpose. Clamp the anchor at the upper end
of the muscle into the drill chuck, and you're ready to go. If
you're making a rod-coiled muscle, you may also want to chuck the
rod, but I'll get to that later.
Self-coiled
muscles are easy from this point onward. You just need to keep the
drill going until the entire working length of the muscle has formed
into a series of secondary loops. Self-coiled muscles will even
“soft set” prior to annealing. When you remove the suspended
weight and let go of the lower end, the muscle should relax a little
and experience a small increase in coil diameter (possibly with a
rather startling spin – watch for flailing paperclips), but it will
hold its basic shape. If you want a rod-coiled muscle, some
additional complexities come into play.
For
a rod-coiled muscle, you need to stop the twisting of Stage 1 before
the muscle begins to self-coil. Doing this consistently is a bit
tricky. You can count the rotations of your drill; the number you
need will vary with the length of nylon you are preparing. You can
time Stage 1 (this works best if you have a partner with good
reaction time, and again, the time will vary with the nylon length).
You can twist until the first secondary self-coil appears, then
reverse your drill just long enough to undo it. Or you can visually
check the candy-stripe pattern formed by the heating wire that might
be part of your muscle; as Stage 1 twisting proceeds, the angle of
the stripe will become more shallow, and you will see more individual
stripes per centimeter of nylon length. When Stage 1 is complete,
unhook the suspended weight from the lower end of your muscle. Hold
on to it or it will
untwist itself, form whiskers, or do something else stupid. Or else
just lift up the weight with your hand, maintaining the existing
tension.
Now
the twisted nylon must be wound around the rod. The best trick I've
found for this is to start out by putting the rod into the drill
chuck with the upper anchor. During Stage 1, it will lie parallel to
the rod and get twisted up, without interacting with the rod much.
When Stage 2 begins, you take the lower end of the nylon and pull it
upward and outward – so that the nylon is now perpendicular to the
rod – and start the drill again. This will automatically wind the
nylon around the rod and give you nice, even coils. The direction
you run the drill at this point determines whether you will get a
homochiral or heterochiral product. Stand in front of the drill and
note which direction the front face of the rotor turns during Stage
1. Supposing
it moves toward your right, and you pull the nylon to your right when
beginning Stage 2, and you keep running the drill in the same
direction, your muscle should be homochiral. If you reverse the
drill, it will be heterochiral.
You can double-check by observing how the Stage 2 coils want to lie
on the rod as they form. If they naturally pull toward each other
and close-pack on the rod as densely as possible, it's homochiral.
If they seem to repel each other, it's heterochiral.
Homochiral muscle, Stage 2 winding. Angle the end downward for best results. |
Once
your muscle is wound, you need to secure it to the rod so it can't
come undone. I clamp the paper clips to the rod with alligator
clips. Again, this whole assembly will be getting very hot, so don't
use anything that can't take it. When you are finished winding your
muscle, I recommend doing a continuity check between the two ends of
the heating wire before proceeding to heat treatment. It can snap in the middle without your noticing.
Now
the muscle must be annealed (heat-treated so it will hold its new
shape). The recommended temperature is about 150º C (300º F), and
the muscle should be kept at this temperature for many minutes. I
bake mine for at least 20 minutes; longer does not hurt. Once you've
twisted your muscle, you may be able to anneal it in place with a
heat gun, but I prefer to put them in my toaster oven. I set the
muscles up in the little pan that came with the oven, such that no
part of the muscle touches the pan's bottom or sides (you wouldn't
want uneven heat transfer to produce irregular results).
Oven
dials are unreliable, so check the actual temperature of your oven
with a kitchen thermometer or calibrate it with sugar to see what
you're really
getting before trying to make muscles. If the oven has a preheat
function, use it, and make sure it's up to temperature before putting
the muscles in; if it doesn't, let the oven soak for a while after
you turn it on. (I wait ten minutes before introducing the pan with
the muscles in it.) When annealing is finished, remove the muscles
from the oven and don't touch them until they are cool.
If
you've made a rod-coiled homochiral muscle, you must now ask yourself
whether you would like to spread the coils. If you recall,
homochiral muscles contract when heated, and they naturally form with
perfectly close-packed coils. But if the coils are all right next to
each other, there's nowhere for the muscle to contract to
… which means it can't work unless it is put under tension.
Depending on your application, you might want there to be some space
between the coils when the muscle is at rest, even if it is not under
load. To achieve this, you can stretch the muscle apart on the rod,
position the coils evenly with your fingernail, and return it to the
oven for a second
annealing phase. I don't recommend trying to spread the coils before
the first annealing phase. The muscle is under fierce tension at
this point and will rebel against you, so you may end up with an
ugly-looking and inefficient monstrosity.
When
annealing and cooling are complete … you have a muscle! Slip it
off the rod (if necessary) and put it to use.
Muscle
usage and control
Muscles
with heating wires can be electrically controlled. After a muscle
has been actuated, it is possible to hold it in position by applying
enough current to keep it stable at some temperature, corresponding
to the desired amount of deflection. Care must be taken to ensure
that this temperature is below that which will permanently damage the
muscle. Too much current applied for too long can result in
overheating, flattening the coils or even melting the nylon clean
through. (Note that if you do cause a muscle to “go flat,” you
can get it back into shape by slipping it onto a rod and annealing it
all over again.)
A
muscle may tolerate a larger current than the holding current for a
short period of time, when it is first beginning to warm up. In
fact, an initial high-current spurt is probably desirable to make the
muscle heat quickly, reducing response time. Thus the control scheme
should focus on providing this initial surge to bring the muscle up
to the holding temperature, without producing a large overshoot that
will damage the muscle. Precise position control almost certainly
requires feedback; I have not attempted this myself, yet.
The
load a muscle can actuate depends on its material and form. In
general, thicker nylon yields stronger muscles; there is also an
inverse relationship between load capacity and rod diameter, for
rod-coiled muscles. All else (e.g. coil spacing, load) being equal,
rod diameter also affects the maximum deformation as a percentage of
the muscle's length, with large-diameter muscles being able to move
farther; thus the diameter of the rod can be used to trade actuation
distance against actuation strength. For a detailed examination of
this trade-off, see Artificial Muscles VI.
Nylon artificial muscle papers, news articles and resources
Nylon artificial muscle papers, news articles and resources
Note:
the Science
paper that started all the hoopla is behind a paywall.
io9: Scientists just created some of the most powerful muscles in existence (This is a good overview of the original discovery)
Txchnologist: Fishing Line Muscles Could Make Superstrong Robots (This article has a bunch of neat GIFs)
Polymer Solutions Blog: Artificial Muscles from Cheap Polymer Fibers? (This article has a few more juicy details from the original paper
than the others do)
Other
nylon artificial muscle hobbyist web sites/video channels:
Barbara
Trost: https://www.youtube.com/watch?v=2OuRX65xbKE
Christian
DEHAIS: https://www.youtube.com/watch?v=vn33DZjERJc
Jason
M. Head: http://robotics.jasonmhead.com/
Lupus
Mechanicus: https://hackaday.io/project/4726-li-s-si-batteries
Attila
Blade: https://www.youtube.com/watch?v=u0UF4xnGF14
Iskanderuse:
https://www.youtube.com/watch?v=NZ9oRwdHWyc
Michael
Picchini: http://monograph.io/micpic/diving-in-to-artificial-muscles
Complete
list of artificial muscle posts on the WriterOfMinds blog:
Position control of fishing line artificial muscles (coiled polymer actuators) from nylon thread
ReplyDeletePaper 9798-99
Author(s): Takeshi Arakawa, Kentaro Takagi, Nagoya Univ. (Japan); Kenji Tahara, Kyushu Univ. (Japan); Kinji Asaka, National Institute of Advanced Industrial Science and Technology (Japan)
Fishing line artificial muscles are fabricated by coiling polymer filaments. We may call such a kind of actuators as coiled polymer actuators (CPAs). In this paper, a CPA is made from Nylon fishing line and Ni-Cr alloy (Nichrome) wire is wound around it. Taking account of the physical principle, two first-order transfer functions are introduced in the proposed model. The parameters of the model are estimated by the system identification. Finally, a 2-DOF PID controller is designed and the position control of the voltage-driven CPA is demonstrated in the experiment.
Why not utilize heating and cooling via an outside source- such as a dual-mode hair-dryer? That way, so long as the source air can be shifted from a cooler source than the wire is used to, both heating and cooling can be expressed.
ReplyDeleteJust a thought.
It's a good idea for demos. In an application, though, if you're trying to use artificial muscles as a motor replacement, adding a fan or blower back into the system to help actuate the muscles kind of defeats the purpose.
DeleteA rather simple question for you, please... Once you have a self-coiled sample, do you release it from the apparatus and allow it release some of the tension before you put it in the oven for annealing? Or do you keep the coiled muscle under tension and anneal it with out allowing it to relax?
ReplyDeleteI release the self-coiled muscles from tension before annealing them, simply because I haven't found a practical way to keep them tensed while they go through the oven.
DeleteI found video with this artificial muscles in action
ReplyDeletehttps://youtu.be/UbvNW7ONiTc
I had tried your method by a making rod-coiled artificial muscle but i couldn't make it. Could you show me a video how to wind an artificial muscle on the rod? Thank you so much.
ReplyDeletePlease Date your posts. Without a date they are much less useful.
ReplyDeletecan someone help me out, I've been trying the mass tension formula for the past 2 hours and cant reach a conclusion that makes sense. Please if anyone can make sense of this that would be awesome.
ReplyDeleteIt's a basic pressure equation: F=pA, force = pressure * area. Since we're told to plug in 17MPa for T, we're working in SI units. The pascal is defined as one newton per square meter, so your area is in square meters. πr^2 is the area of a circle, so r is in meters. One newton is the force needed to accelerate one kilogram of mass at the rate of one meter per second squared in direction of the applied force, so F=mg, force=mass*the acceleration of gravity (9.81m/s^2). We know the force and want the mass, so the equation was rearranged to m=F/g.
DeleteSee my other comment addressed to you for an example.
Jakson, r is the radius in meters. T is in Pascals (17MPa=17,000,000Pa). g (gravity) is 9.81. m is in kilograms.
ReplyDeleteFor 381μm fishing line, it would be
m = π * (0.000381/2)^2 * 17,000,000 / 9.81 = 0.198
198 grams agrees with the mass given in the table in Nylon Fishing Line Artificial Muscles II.
The show stopper is efficiency. Haines, et al, tell us, "The
ReplyDeletemaximum specific work during contraction was
2.48 kJ/kg for the C = 1.1 nylon 6,6 muscle of Fig. 2C (1)." The temperature range used was 20° to 120°C, giving ΔT=100. The specific heat of Nylon 6,6 is 1.7J/g°C, so
2.48/(1.7*100)=1.46%,
which is quite close to the maximum efficiency of 1.08% attained by Haines, et al.
RoboChicken will need a car battery, which makes him a bit too heavy to fly.
This comment has been removed by the author.
ReplyDelete