Pneumatic Artificial Muscles

McKibben air muscles were invented for orthotics in the 1950s. They have the advantages of being lightweight, easy to fabricate, are self limiting (have a maximum contraction) and have load-length curves similar to human muscle. The muscles consist of an inflatable inner tube/bladder inside a braided mesh, clamped at the ends. When the inner bladder is pressurized and expands, the geometry of the mesh acts like a scissor linkage and translates this radial expansion into linear contraction.

Standard McKibbens contract in a linear motion up to a maximum of typically 25%, though different materials and construction may yield contractions around 40% . Though they can technically be designed to lengthen as well, this is not useful as the soft muscles buckle.

Some of the information contained in this web site includes intellectual property covered by both issued and pending patent applications. It is intended solely for research, educational and scholarly purposes by not-for-profit research organizations. If you have interest in specific technologies for commercial applications, please contact us here.

Design

Actuator components

  • Elastomeric inner bladder – inflates when pressurized with fluid. Thinner bladders make the actuators more efficient, as less energy goes into stretching the bladder material, but if the bladder is too thin it will tear/pop easily, especially since it rubs against the mesh. The friction between the bladder and the mesh also contributes to actuator hysteresis.
  • Braided outer mesh – constrains bladder motion, translates radial expansion of bladder into linear contraction. Utilizing asymmetric meshes can enable other motions such as bending or twisting, but require custom fabrication.
  • End clamps – hold everything together, prevent slip of mesh relative to inner bladder, provide attachment point to mount actuator

Variation: Geometry

  • Larger McKibbens with larger diameters can apply larger forces, but also need more air (so slower actuation for a given pump).
  • Because actuator contraction is limited by a fixed ratio, to get increased travel of the actuator end, the original length of the actuator must be increased.

Variation: Bladder stiffness

The stiffness of the inner bladder can be changed by varying the material, using materials with different durometers and elasticities, or by varying the geometry, i.e. changing the wall thickness. Using the testing methods described here,  the following effects of stiffness on actuator behavior were observed:

Isometric testing shows that the force-pressure relationship is independent of stiffness. As actuators were not allowed to contract, little energy goes into stretching the bladder material, and so the stiffness of the bladder material does not have a significant effect.

Stiffer actuators are slower to reach a given amount of force.

If pressure is held constant, actuators which are less stiff can reach higher contraction amounts and apply more force.

Variation: Initial mesh braid angle

There is a wide variety of braided meshes available from manufacturers, with different diameters, materials, fiber sizes, etc. One important characteristic of braided meshes to consider during mesh selection is the initial braid angle.

The braid angle is the angle formed between the longitudinal axis of the mesh and the mesh fibers.

The braid angle can be measured by examining the mesh under a microscope:

In general, lower initial braid angles are preferable. There is a neutral/"magic" angle of 54°, which can be derived analytically, at which no more axial contraction or radial expansion of the mesh can occur. Mesh braid angles typically start below this neutral angle, and move toward it as the actuator contracts. Therefore, meshes that start at a lower angle will have a longer "stroke length" and contraction percentage, with corresponding higher forces.

Varying braid angle has the following effects:

In dynamic tests, actuators with lower braid angles achieved higher forces, and reached those maximum forces more quickly.

In quasistatic isometric tests, actuators with lower braid angles achieved higher force for a given pressure.

In constant pressure tests, actuators with lower braid angles achieved higher forces and contraction amounts.

Fabrication

There are many different methods of fabricating McKibben actuators, depending on the desired specifications and intended application. However, the general fabrication process is as follows:

  1. Get inner tubing (can be off-the-shelf or custom-molded) and cut to length
  2. Cut mesh to length and seal one end
  3. Optional: Modify end braid angles (helps prevent ends from fraying)
  4. Pull mesh over inner tubing
  5. Add air line and seal/clamp
  6. Optional: Coat actuator with outer layer of elastomer (for faster deflation)

Example Fabrication

The actuator made in the below guide can be made in under 10 minutes using entirely off-the-shelf parts. Though this is one specific fabrication process, it demonstrates the more general process outlined above.

Materials

  • Party balloon (long thin kind used for making balloon animals)
    • This will be used for the inner bladder, other types of elastic tubing would work also
  • Braided mesh sleeving
  • Barbed connector (use one that fits your pressurized air source for easy attachment later)
  • Zip tie
  • Lighter
  • Nut, washer, or similar tool with a hole diameter smaller than the braided mesh

Assembly

Cut a piece of mesh 4-5 inches long. Cut a piece of party balloon about the same size.
Gather the strands of one end of the mesh and hold them together with a tool, leaving 3-4 mm sticking out the end. Several tools can be used, i.e. pliers, a metal fixture, etc. Here we use a nut.

Hold a flame very close until the strands melt and fuse into one clump.

Remove the nut. The end should not fray even when the mesh is compressed.

Put the balloon in the mesh. Insert the balloon 1-2 cm without fraying the mesh too much, then use the following “inchworming” process [Video]:

  1. With your right hand, pinch the balloon + open end of the mesh, and with your left, pinch the mesh a few cm in front of the balloon tip.
  2. Bring your fingers together, until your left hand is almost touching the balloon. The mesh will expand radially.
  3. Scoot your left pinch forward so that you are pinching the front tip of the balloon.
  4. Release your other hand. The mesh will slide backwards over the rest of the balloon.
  5. Repeat steps 1-4 until the balloon is all the way in the mesh.
Trim away any frayed mesh and trim the balloon to match the mesh length.

Insert the barbed connector, with the larger end going into the balloon.

Use the zip-tie to secure the mesh and balloon onto the barbed connector, clamping onto the narrow section immediately after the barb.

Actuation

Trim the excess zip tie ends, then connect the actuator to a pressurized air source like a hand pump or wall valve.

Pressurize the actuator. It should contract about 25% of its original length. Be careful not to over-inflate as it may pop.

 

If your actuator leaks, you can re-do the zip-tie clamping after adding some padding around the barbed connector, or use thread wrapping instead of the zip-tie.

Inner tubing

There are various options for the inner tubing used in making McKibben actuators. Below are some rough guidelines on the different options and their advantages/disadvantages.

Balloons

Party balloons are good for quick prototyping as they are easy to buy and modify. However, they are not very reliable. Also, as they are made of latex, they cannot be cast in silicone (latex inhibits silicone curing) and can trigger allergic reactions in some people, which is a limitation when it comes to medical applications.

Another option is custom-made balloons of the kind used in medical procedures (i.e. stent placement). These are more reliable, but also much more expensive.

Bought tubing

Purchasing rubber tubing makes fabrication much easier since it's just off the shelf.

However, there are fewer options -- for geometry, selection is limited to standard diameters/sizes, and typically it is difficult to get low-modulus tubing (the softest tubings on McMaster are still too stiff for use in McKibbens). This means that actuators made with bought tubings generally require higher pressure.

When selecting off-the-shelf tubing/choosing a supplier, it's important to look at shore scale of durometer, wall thickness, and modulus at the % strain range that you will be working in (generally <100%).

Some supplier options:

Self-cast tubing

With self-casting, it is possible to make tubing with much lower durometers than is commercially available, which enables very low threshold pressures. It is also possible to make the whole actuator monolithic (e.g. end caps) and lends itself to casting in a matrix of the same material. However, self-cast tubing is time-consuming to make and has lower yields (can get up to 80% with experience).

There are several methods of self-casting tubing, two of which are described in this documentation:

1) 3-D printed mold in 2 halves

Pros:

  • Can make tubing of any ID & OD desired
  • Demolding is easy

Cons:

  • 3-D printing can get expensive
  • Molds are hard to clean and material is fragile (alignment posts tend to snap off)
  • Messy, lots of leakage, wasted elastomer
  • Long rods bend, causing off-center inner cavity and non-uniform wall thickness
  • Resulting tubing has “web” on side

2) Molds made using off-the-shelf pipes

Pros:

  • Works well for small, thin-walled tubing
  • Easy to make many in parallel
  • Mold materials are cheap

Cons:

  • Does not work well for longer tubes
  • Making the molds requires more manual work (compared to 3D printing)
  • Tube outer diameter determined by the inner diameter of off-the-shelf plastic tubing, which comes in discrete sizes (so limited customization)

Custom tubing (method 1)

Materials

Ecoflex silicone rubber (2-part) vacuum bagging tape mold release
teflon tape thin metal rod to mold inner cavity

Tools

Thinky mixer Thinky mixing cup syringe
mass scale clamps

Mold components

base stand
mold body (2 pieces) alignment piece for top of rod reservoir

Prepare molds

Once 3-D printed, the molds should be thoroughly cleaned as residual support materials may cause cure inhibition. Methods vary between plastics/printers, but here are some suggestions:

  • Print with a glossy finish, then leave for 4hrs in a 70°C oven
  • Clean with soap and water, then IPA, then dry in the oven
  • Do a sacrificial molding out of Ecoflex

Assemble molds

In a fume hood or other well-ventilated area, spray the inside surfaces of the mold and the inner rod with mold release (hold the can about 20cm above the molds). Let dry for a few minutes.

Place the two halves of the mold together, combine it with the reservoir/alignment piece and base stand, and seal the seams by wrapping them with Teflon tape.

[Video: Teflon wrap mold]

Add several clamps to hold the mold halves tightly together. Stagger the clamps so that the mold stays balanced upright.

[Video: Clamp mold]

Make sure the inner rod is straight (roll it around on a flat surface to check), then insert it into the mold. It should lock into the corresponding hole in the base piece.

Mix elastomer

Make the elastomer for molding the inner tubing. The amount will vary depending on the size of your actuator, but make 20g more than you need in case of leaks.

  • For a low stiffness actuator, use Ecoflex 00-30, which is mixed in a 1:1 ratio.
  • For a high stiffness actuator, use Elastosil M4601, which is mixed in a 1:9 ratio.

Weigh out and combine the two parts of the selected elastomer in the appropriate ratio, and place in a Thinky mixer.

Fill syringe

Move any PTFE tape out of the way to expose the injection hole on the mold. Depending on the nozzle of the syringe and the dimensions of the injection hole, you may need an adapter, i.e. a Luer fitting or a snipped off piece of a disposable pipette, so that it fits snugly into the injection hole.

Luer adapter pipette adapter

Fill the syringe with Ecoflex. However, to avoid air bubbles and since the material is very viscous, the syringe should be filled from the top. Remove the plunger, block the nozzle with a finger, then slowly pour the mixed Ecoflex into the syringe until it is almost full.

Hold the syringe over a cup (for spillage) then re-insert the plunger, pushing it in until it locks with a popping noise.

[Video: Fill syringe]

Inject

Slowly inject the Ecoflex into the mold until some of it enters the reservoir at the top of the mold. There will probably be some leakage, but if it isn’t major, continue – otherwise, plug with more clamps/tape.

[Video: Inject]

Remove the syringe, then plug the injection hole with clay.

Cure

Place the mold in a pressure oven set at ~75 psi for at least 1 hour, and let it cure for 4 hours total before demolding.

Demold

Pry open the molds. You may need to use pliers on the side notches to initially force the mold open. Gently remove the molded tubing. Inspect it for holes or sections where the wall thickness is very uneven, and cut away these parts.

Plug end

Mix a small amount of Ecoflex.

For these actuators, the inner channel is small enough that you can simply dip the end into elastomer, and capillary action will pull some elastomer into the tube to create a plug.

Wipe off excess elastomer from the outside of the tube, then heatgun for a few seconds until the material is no longer tacky.

Custom tubing (method 2)

Mold fabrication

  • Cut a 90mm length of clear plastic tubing (i.e. polycarbonate) that has 1/8” inner diameter.
  • With a #32 drill, make 2 holes, each ~5 mm away from the tube end. These will be the injection port/exhaust holes.
  • Sand/file the cut ends to smooth them out.
  • Use a blade to slice away any burrs.


Cut a short length of 1/16” dia. metal rod, a few mm longer than the plastic tubing.

File the rod ends until they are smooth, so that they will not stab/tear the molded tubing later on.

Wrap tape around each end of the rod, so that it kind of looks like a little barbell. These taped ends simultaneously center the rod within the mold, and plug the open holes. Electrical tape is good because it is slightly compressible.

Mold assembly


Remove the plugs and try fitting them into the end holes. They should fit snugly – if they don’t, adjust the length of tape wrapped around the rod.

Once the plugs are properly sized, insert the rod into the tube, then insert a plug on either end, centering the rod.

It is okay to partially obstruct the injection/exhaust holes, but don’t completely block them.

Inject

Fill the syringe (procedure shown in other method)

Align the mold(s) so that the injection/exhaust holes are facing up. You can place the molds on piece of tape so they stay in this orientation and don’t roll around.

Slowly inject elastomer into each mold until it slightly overflows out the exhaust port.

[Video: Inject]

Cure


Place in a pressure oven set at ~75 psi for at least 1 hour, and let it cure for 4 hours total before demolding.

Demold

Twist off/pull out the tape plugs.

Grip the inner rod using pliers, and pull it out slowly.

[Video: Remove plugs and rod]


Grab one end of the molded part. You might have to fish out the end with a rod or other thin tool.

Pull out the tubing, going slowly so that the elastomer unsticks itself as you go, without stretching the tubing too much. Injecting IPA helps the tubing come out much more easily.

[Video: Demold tubing]

Plug end

Mix a small amount of Ecoflex.

For these actuators, the inner channel is small enough that you can simply dip the end into elastomer, and capillary action will pull some elastomer into the tube to create a plug.

Wipe off excess elastomer from the outside of the tube, then heatgun for a few seconds until the material is no longer tacky.

Preparing braided mesh

Materials

Braided mesh sleeving (i.e. TechFlex)
1/8” dia, cut to 75mm long piece
Metal support rod
(slightly larger diameter than mesh)
Lighter

Lock end

Push the mesh onto the support rod, leaving a few mm hanging off the end.

Hold a flame 2-3 mm away from the mesh, rotating it slowly, until the loose strands melt and fuse into a ring.

[Video: Melt ring]

Remove the mesh from the rod.

Close other end

Gather the strands of the other end into a bundle and hold them together with a tool. There are several ways to do this – i.e. a M2.5 lock nut is used here – such as clamping forceps, a metal fixture with a hole in it, a washer, etc.

[Video: Add nut]


Hold a flame 2-3mm away until the ends melt and fuse, into one clump this time.

[Video: Close end]

Check

Check that the ends are secure by compressing the mesh against a flat surface. It should not fray.

[Video: Test end]

Add mesh and finish

Slide on braided mesh

The next step is to insert the molded tube into the mesh. Because  the tube can only move through the mesh when it is compressed/expanded, you need to use an “inchworming” process [Video: Insert Tube]:

  1. Push the tube forward, making the mesh around it linearly compress/radially expand.
  2. Grip the forward end of the tube, then release the back end. The tube will be stretched, with its front end having advanced relative to the mesh..
  3. Expand the back end of the mesh so that the back end of the tube can move forward and catch up with the front end of the mesh.
  4. Repeat.

It takes some practice and a lot of repetition. Continue until the tube is almost at the closed end of the mesh.

Insert air hose

Insert the air hose into the molded tubing – there should be about 10mm of overlap. Make sure that it is centered and that the elastomer is not stretched unevenly on one side (like in the first image).

Use the “inchworm” method again to insert the overlap region into the mesh.

[Video: Insert air hose]

Secure hose and mesh

Tie the mesh and tube onto the air hose using nylon thread. Make 3-4 wraps around the center of the overlap, then secure with a knot (square knot works well).

Tighten the knot with pliers. Trim off the excess thread.

[Video: Tie on mesh]

Coat

Coat the actuator with elastomer, either dipping it or coating it with a brush or gloved finger. Wipe off excess elastomer so that only a thin layer remains, and keep rotating the actuator to keep the coating even all around.

[Video: Coat actuator]

Heatgun to cure.

Modeling

Analytical

There are several existing analytical models for McKibbens/PAMs, which vary in complexity as well as closeness to experimental results. The most widely cited model is by Chou and Hannaford, a static model assuming an ideal actuator that is cylindrical (no tapering at the ends) and ignores friction and bladder wall thickness/elasticity. Using conservation of energy principles and braid geometry, the following relationship between force, pressure, and braid angle is derived:

Where D0 is the mesh diameter when braid angle is 90°. Setting F=0 and solving for braid angle yields 54.7 degrees, the neutral/“magic” angle at which the maximal contraction is reached.

The Chou and Hannaford model has been found to deviate up to 15% from experimental results. Some models take into account factors such as friction between braid fibers, reducing the deviation from experimental results to 5% at the cost of greatly increased complexity.

Finite Element

To model the response of the actuators to an increase in pressure, without  the need for a detailed model of the braided mesh, temperature and orthotropic coefficients of thermal expansion can be used to simulate the anisotropic strain response of these actuators.

This is useful for modeling devices where the actuators are integrated into other materials, such as in the cardiac simulator case study where actuators were arranged and embedded inside an elastomeric matrix. Actuators were assigned an experimentally derived modulus of 1.78 MPa and orthotropic thermal expansion coefficients according to experimentally derived strains that were negative in the longitudinal direction and positive in the radial direction for a positive change in pressure.

Testing

There are two general types of tests to evaluate McKibben actuators:

  1. Actuator length held constant (isometric), and the force measured as the actuator is pressurized
  2. Actuator pressure held constant, and the force-contraction relationship measured

Both tests are carried out using tensile testing machines (i.e. Instron) which are typically used for evaluating material properties.

Isometric test

In isometric testing, actuators are clamped at both ends, with one clamp having a load cell to measure forces. The clamps are fixed, holding actuator length constant while the actuators are pressurized. Pressure is measured with a pressure transducer attached to the actuator air supply line.

Both quasistatic and dynamic tests can be performed:

  • Quasistatic test: pressure is ramped from 0kPa to some maximum pressure and back using a pressure regulator, over the course of ~1 minute (slow).
  • Dynamic test: a solenoid valve is opened rapidly (millisecond time scale) to quickly deliver air from a pressurized chamber to the  actuator.

Quasistatic testing yields a force-pressure curve. McKibben actuators show a very linear force-pressure relationship, as seen in the below graph. The vertical line at the low pressure end of the graph is due to the fact that the pressure transducer was not sensitive enough.

Dynamic testing yields force-time curves and can be used to evaluate how quickly actuators respond to the pressure input.

Constant pressure test

Once again, the actuators are clamped at both ends, with one end attached to a load cell. However, this time one end is allowed to move. While pressure is held constant by an accumulator, actuators are allowed to contract until no load is measured at the supports, and then are stretched back to their original length. The displacement is measured and used to calculate percent contraction (displacement/initial active length).

As expected, force decreases with increasing contraction, regardless of the various constant pressures at which the tests were conducted.

Failure testing

It is important to determine the pressure limits of each actuator design, especially when it comes to the safety of using these actuators in various applications. A simple method for this is to gradually pressurize the actuator (with no load attached) until it fails. However, since failure happens very rapidly, it is helpful to film the actuator with a pressure readout (of a pressure sensor in-line with the actuator) placed next to it.

Fatigue testing can also be conducted by using an automated system like the control board to cycle the actuator continuously between relaxation and a set contraction amount while tracking the number of cycles. Depending on construction, McKibbens can last anywhere between a few hundred and a few million cycles.

Case Studies

Cardiac Simulator

This case study describes the use of McKibben actuators embedded in a soft elastomeric matrix to mimic heart muscles. 

Soft Prosthetic Hand for Amputees

This case study describes the design and construction of a soft prosthetic hand for amputees designed by a group of students in the Global Immersion Summer Program in India. The thumb is actuated by three pneumatic artificial muscles.

Cardiac Simulator

Clinical need

Existing functional simulators are generally passively driven by flow, the heart walls do not actively contract, and they don’t simulate 3-D twisting motion (with the exception of the Chamberlain surgical trainer) or allow simulation of pathological motion. This limits the clinical relevance of device testing on these test-beds.

Solution

Inspired by biological muscle, where contractile elements are arranged in a soft matrix, soft custom-molded McKibben actuators were fabricated then embedded in a soft elastomeric matrix with material properties close to physiological tissue.

Initial models used simple configurations of actuators in elastomer matrix. Basic geometric properties such as actuator spacing and matrix width/length were varied, and the force and strain behaviors of these models when the actuators are pressurized were observed via image tracking of markers and a tensile test machine.

Once these simple units were characterized, the same principles were applied to the helical arrangement of muscle fibers in an actual heart.

EM trackers were placed on molded alignment features on the simulated ventricle to evaluate its motion. When all actuators were pressurized, apical and basal rotation of the ventricle matching the range of clinical values was observed.  In addition, the effect of deactivating select actuators on the overall motion of the ventricle was observed. This mimics the effect of a section of heart wall losing its function via myocardial infarct or other cardiac diseases and simulates the subsequent pathological motion.

Soft Robotic Prosthetic Hand for Amputees

This case study describes the design and construction of a soft prosthetic hand for amputees designed by a group of students in the Global Immersion Summer Program in India.

 

Clinical Need

Diabetes, peripheral artery disease and trauma cause hundreds of thousands of upper limb amputations worldwide per year. For amputees, restoring the utility of a missing hand is a major factor in being able to do activities of daily living (ADLs) like eating, bathing and dressing themselves. Gaining the ability to do these basic tasks is extremely helpful to the patient achieving a higher quality of life. Current prosthetics including body powered and myoeletric devices give a lot of this function back to the patient, but can be tiring to use, expensive, stiff, or have a limited range of motion.

 

Solution

The team’s solution consists of 4 fiber reinforced bending actuators that mimics the index, middle, ring and little fingers. The thumb is an aluminum mechanism actuated by pneumatic artificial muscles.

The overall goals of the project were to design a soft prosthetic hand that was: cost effective, aesthetic, provided sufficient grip strength for ADLs, low maintenance, and light weight. The design of the hand was broken up into 4 separate modules: Thumb, Palm, Fingers, and Control. These 4 modules were determined to need to come together to into a few potential grip patterns that would be useful to complete ADLs, including pinching, open palm, pointing and a power grip.

The thumb design needed to be rigid and able to transmit forces in multiple different orientations for the different grip patterns. To achieve these grip patterns, the thumb was designed as a hinged aluminum mechanism.

As opposed to the four fingers of the hand, which only need to bend in one plane, the thumb needs multiple degrees of freedom. To achieve this range of motion, the thumb was actuated by three pneumatic artificial muscles in different orientations. These actuators were controlled by a version of the Open Source Control Board described on this site.

The following video shows how the different actuators affect the movement of the thumb:

For more information on the fingers of the hand, visit the case study page in the Fiber Reinforced Actuators section.