Robot Design - Flexinol and other Nitinol Muscle Wires
With its unique ability to contract on demand, Muscle Wire (or more generically, shape memory actuator wire) presents many intriguing possibilities for robotics. Actuator wire is made from an alloy called Nitinol, and marketed under a number of brand names including Flexinol, BioMetal, and Muscle Wires. Nitinol actuator wires are able to contract with significant force, and can be useful in many applications where a servo motor or solenoid might be considered.
Flexinol and Muscle Wires are both trademarks of the California company Dynalloy. Flexinol is referenced in many books and research articles, and has been specified in a number of student and hobby-level robotics projects. Stiquito is a small Flexinol-based hexapod created in the early 1990s by Jonathan Mills at Indiana University. Mills and co-author James Conrad at UNC Charlotte published a number of articles and books on Stiquito, and Stiquito kits are still available today. Perhaps most famously, Flexinol was used to operate a dust sensor installed on the Sojourner rover which landed on Mars as part of the Mars Pathfinder mission July 4, 1997.
Nitinol is one of several alloys which demonstrate the shaped memory effect. When cool these alloys can be bent or stretched, and then when heated the alloy returns to its original shape. In most robotics applications the heat required for this transition is generated by conducting electricity through the wire. Actuator wire will contract by about 3-7% of its length in the transition. The contraction might seem small, but it is accompanied by such force that it can easily be applied through levers and linkages to achieve much larger movements. Nitinol returns to its original shape with a pulling force of about 25,000 pounds per square inch. An actuator wire with a thickness of just 1/100 inch can lift 2 pounds.
The properties of an actuator wire are highly dependent on the manufacture of the Nitinol and the preparation of the wire. Nitinol can be formulated to transition across a wide range of temperatures. When used in medical applications it will frequently be active in the range of human body temperature. LiveWire is a Nitinol product most often used as a novelty or educational item to demonstrate the shape memory effect - it has a low enough transition temperature to react in a glass of hot water. When intended to be an actuator, Nitinol is usually formulated to transition at a temperature achievable with electric current. Flexinol is available in a both low temperature and high temperature formulations to support different applications. Remember that unprepared Nitinol is not an actuator; when heat is applied the only result will be that it gets hotter.
The thickness (or gauge) of the actuator wire will dictate the maximum force that can be applied. Gauge also dictates the wire's resistance to electric current, the available current required to reach transition, and the time the wire will require to heat and cool.
The potential advantages to working with Nitinol actuators are significant. Solutions can be physically small when compared to a servo or solenoid, and weight considerations are often trivial. Additionally, because the activation mechanism is heat rather than magnetism, these solutions do not introduce magnetic noise to a circuit - they are in essence solid state devices.
Of course there is no free lunch. Designing with Nitinol brings its own set of challenges. Working with the thin, tough wires can be physically difficult, and a substantial current can be required to achieve transition. Perhaps the greatest challenge lies in the fact that actuator wire must be stretched back into its lengthened state with each activation cycle. This stretching requires that a counter-balancing force be engineered into the project. Much of the remainder of this article will be concerned with addressing these challenges.
To be used as an actuator, Flexinol must be attached both to the mechanical elements where it will cause movement, as well as to the electric circuit that will energize it. Crimping the wire is the only viable option. The thickness of the wire changes slightly as it goes through transition and therefore solder or glue solutions will ultimately fail. Although it is possible to buy Flexinol pre-crimped, in most cases the hobbyist or student must crimp the wires themselves. The hair-thin nature of most Flexinol gauges makes crimping difficult, but it is certainly possible to get a good crimp with hand tools. There are many potential approaches. In the original Stiquito plans, Mills ties a small, loose knot in the wire and crimps with a section of 1/16 inch aluminum tubing. Roger Gilbertson, author of the Muscle Wires Project Book, recommends the N-scale rail joiners which are used to connect model railroad conductive track. Many builders will use standard solderless terminals from other electronics applications. A good and simple solution for a temporary crimp that can be repositioned is to use standard machine screws and hex nuts.
Flexinol in most applications uses electricity to generate the internal heat needed for transition. Proper control of the current is essential for satisfactory results. Too little current and the wire will fail to contract. Too much current and the wire will overheat, becoming stressed and losing its shape memory properties. In between too little and too much, variations in supplied current will affect the heating and cooling times. In the table reproduced below, Dynalloy provides some guidelines as to how much current is required to cause a Flexinol wire of a given gauge to contract in one second. Also provided is the approximate electrical resistance of the various gauges. Note that Flexinol has fairly high resistance compared with copper wire and other common conductors. In some applications it is possible to activate the wire without additional resistance in the circuit. A 0.005" diameter wire has a resistance of about 1.9 ohms per inch and needs about 320mA to reach its transition temperature - to heat a 5-inch length of 0.005" Flexinol will require about 3 volts which could come directly from two standard AA batteries. (Volts = Resistance in ohms * Current in amps.) Of course Flexinol is hungry for current and it would be easy to overload a wire. If above there were 2 inches of wire rather than 5, that same 3 volts would deliver 1.58 amps and likely destroy the Flexinol.
With wires .006" or thinner, Dynalloy says that the contraction current indicated in the chart can be applied continuously. However thicker wires will not be able to dissipate the heat quickly enough and it will be necessary to make some adjustments. Part of the solution rests upon how fast the response from the wire is required to be. If a longer cycle time is acceptable, then lower current can be supplied. To maintain a one second response, current can be pulsed in short bursts and then removed. Additionally, heat sinking and other cooling countermeasures can be brought into play. Of course, a combination of these solutions might be in order. Success can be determined experimentally. If the Flexinol begins to contract immediately after current is removed, then it is probably at about the correct temperature. If on the other hand, the wire does not begin to contract immediately after current is removed, then it is probably running too hot and in danger of being damaged.
For obvious reasons, the added complexities of working with the thicker wires make the lighter gauges particularly attractive for small, hobby robotics projects. Actuators made with .006" wire and below can easily be controlled by basic microcontroller techniques and powered by simple transistor arrays such as the familiar Darlington transistor ICs ULN2003 or ULN2803. (See Flexinol Control Circuit Using PIC 16F690 and ULN2003A for more information.) To achieve greater pull from the thinner wires, more actuators can be used.
Flexinol does not automatically return to its lengthened state when cool. A force must be applied to the wire to pull it back to length. The force required for this recovery is about 1/5 the force with which the Flexinol contracts. In some applications it is possible to use the natural motion of the mechanism itself to provide the recovery force. For example, in the case of an arm that lifts a dead weight, when power is removed and the Flexinol cools, the weight of the payload will stretch the wire back to its extended length. Likewise, a robot leg that lifts the machine itself can use the weight of the robot to refresh the Flexinol.
In most cases it is necessary to engineer a biasing force into the mechanism. Potential bias forces include gravity with a counter weight or another length of Flexinol. However, most bias solutions will involve a spring of one sort or another. A spring in this context is any elastic device that will regain its original shape after being compressed or extended. A couple of convenient and readily accessible sources of springs for experimentation include rubber bands and the compression springs found in retractable ball point pens.
Music wire (also called piano wire) can be used to fabricate your own springs. This strong and refined steel wire is inexpensive and can be bought in many gauges from hobby suppliers. Alternatively, guitar strings are made from the same material. Bear in mind that the wire does not need to be coiled to act as a spring, a simple bend in the wire can be enough to provide a very functional spring effect. Experiment with different lengths, gauges, and angles in the bend to achieve the correct properties. If you do choose to wind your own coils then be warned - you must use proper tools and wear eye protection even when working with very light gauges of music wire. If you try to wrap a coil using only pliers and a dowel you will almost certainly hurt yourself.
When designing the mechanism care must be taken to ensure the the Flexinol is free to move along its entire range of contraction. If the spring is too strong, or if there is a physical obstruction to movement, then the Flexinol is likely to be stressed and may lose some or all of its shape memory properties. The ideal bias value will of course depend on the application. Some designs might call for a bias as close to the required refresh force as possible to provide as much work energy as possible from the actuator. Other designs might call for a relatively powerful spring in the bias which will perform some of the work required.
The need for a bias force adds a dimension to robot design which is likely new for many hobbyists. Different elastic materials and different spring implementations will produce dramatically different results and some study of the field is necessary to develop sufficient understanding. As with learning anything new, incorporating bias into your designs provides an opportunity for creative problem solving. Consider the very simple robotic gripper illustrated nearby. It uses Flexinol and a rubber band around a pivot. The rubber band functions as an extension spring keeping tension on the Flexinol and keeping the gripper in a "normally closed" state. The device only opens when power is applied. The configuration allows the gripper to grasp objects of various sizes using power derived from the spring. If it were the Flexinol pulling the gripper to the closed position then it would need to go to a fairly precise location and could only grasp objects of a specific size, or else risk damage to the object or the Flexinol.
Another aspect of this simple device worth noting is the use of leverage. The perhaps more familiar use of the lever and fulcrum (pivot) is to allow a small force moving over a large distance to create a large force moving over a short distance. However, it is equally valid to use a lever to do the opposite and enable a large force moving over a short distance to create a small force moving over a large distance. This second application is exactly what happens in the operation of the gripper.
Flexinol is a great addition to your robotic bag of tricks. Work to integrate this unique "muscle wire" into your next project and start designing robots which are lighter, less expensive, and more natural in their movements.
About the Author: Ralph Heymsfeld is the founder and principal of Sully Station Solutions. His interests include artificial intelligence, machine learning, robotics and embedded systems. His writings on these on other diverse topics appear regularly here and across the Internet.