Is this a case of a good design gone wrong in the build phase? Or is this DIY prosthetic arm a poor design from the get-go? Either way, [Will Donaldson] needs some feedback, and Hackaday is just the right place for that.
Up front, we’ll say kudos to [Will] for having the guts to post a build that’s less than successful. And we’ll stipulate that when it comes to fully articulated prosthetic hands, it’s easy to fail. His design is ambitious, with an opposable thumb, fingers with three phalanges each, a ball and socket wrist, and internal servos driving everything. It’s also aesthetically pleasing, with a little bit of an I, Robot meets Stormtrooper look.
But [Will]’s build was plagued with print problems from the start, possibly due to the complex nature of the bosses and guides within the palm for all the finger servos. Bad prints led to creaky joints and broken servos. The servos themselves were a source of consternation, modified as they were for continuous rotation and broken apart for remotely mounting their pots in the hand’s knuckles. The video below relates the tale of woe.
There’s a lot to admire with [Will]’s build, but it certainly still has its issues. He’s almost to the point of other more successful DIY hand builds but just needs a little help. What say you in the comments line?
31 thoughts on “Fail Of The Week: Good Prosthetic Hand Design Goes Bad”
A few suggestions:
1. Don’t give up. I applaud your courage for sharing your story.
2. Start simpler, with just one finger, until the mechanical issues for a single finger are fully resolved.
3. Once you have repeatable and reliable control over one finger and motor, you can re-use the final finger design with successive palm iterations that increase the finger count.
4. If you are going to disable the pots from the servos anyway, you may want to consider using micro steppers with integrated gearboxes (such as those from Faulhaber, for example) instead. These are not cheap, which is another good reason to start with fewer fingers.
5. Homing could be done with optical sensors, limit switches, or (if the mechanics are sturdy enough) simply stepping the motors until they stall. In practice, this means you would command move where the number of steps is slightly more than the worst case full travel step count for each finger.
That design looks overengineered, but made with cheap materials/tools.
Sometimes, overengineering can be used to secure the success of a design, but only when the design is implemented using high precision tools and high quality materials. Otherwise, for a DIY stile, with cheap/improvised parts/tools, a design must be dead-simple. Overengineering can not save the day for this situation.
I forgot to say that the project, in my opinion, is not a fail.
Non working iterations of a design are always expected when making something new. With all the lessons learned, aha moments and the fun of building, it was definitely not a fail. Now you know. Do it better!
Yeah this was my impression. That’s not a fail, that’s v1.
The “failure” (it’s a great job indeed) is only due to those crappy TowerPro clones. I suggest to replace them with digital servo with metal gears like these: https://www.aliexpress.com/item/Free-shipping-20sets-Metal-gear-Digital-MG90S-9g-Servo-Upgraded-SG90-For-Rc-Helicopter-plane-boat/32555610718.html
Lean methodology: Try, try again. I cannot comment on the build quality but I will say, don’t give up. Build, measure, learn. Do cheap experiments to validate hypotheses that allow you to pivot quickly and fail cheaply. This will give yourself a long enough runway toward liftoff and success.
1. do an ugly proof of concept in cheap and quick&easy handling material first. Once you are confident you nailed it, go for aesthetics. If there is more than one innovation in your build, make a separate proof of concept model for each one.
2. ditto for scaling, even when you get to (aesthetic) design phase: big models of small things, small models of big things. Help yourself by scaling things to size which is easy to handle. When you figured it all out, next step is to bring it down or up to scale.
3. when prototyping your first iteration of a “real thing”, it is good to work on a small series (2-5) of identical design, if it is affordable.
4. using a “death grip” glue for putting together a prototype assured that you couldn’t amend it any further, nor repair anything which broke inside.
4a. keeping the prototype form “skeletal” – leaving as wide openings as possible would allow access to inner parts even when it is assembled by gluing.
I always use hot glue for prototypes because if you use it correctly it creates a strong bond, but you can always get it off of plastic parts if you need to.
Stupidity using the arduino k’s output 5V to drive servos, rookie error.
More like lack of sufficient experience/knowledge. Anyone who learns from their experiences is not stupid. Maybe you’d care to recommend a practical non-rookie solution?
Best comment on this article.
You tell him!
To fix the thumb weakness, make the rotating joint supported with a ball bearing or two, or a single, small, double row ball bearing instead of depending on the wee little servo shaft to bear the full weight and forces on the thumb.
A Prosthetic Hand is generally classified into two categories:
1. Functional to reduce task complexity
2. Aesthetics to reduce social anxiety
When you begin building something you start by defining the functional context to limit the scope of your project.
Simply defining a Finish line (harder than most would initially believe):
a.) We want to build simple tool holders to assist rapid effective employment of amputees
Limiting scope (i.e. a small tangible step in the right direction):
b.) Simple adapters for existing hand tools people already use in a garage
Defining a functional context
c.) it will be attached to a person’s arm
Identifying initial design criteria given a., b., and c.
d.) It must be lightweight, economical/disposable, easy to build by a hobbyist, and comfortable for users
Design draft iteration given d:
i.) Adapter for off-the-shelf electric impact wrench, drill, and saws-all
ii.) Adapter for Pneumatic tools like Air Nippers, Air Crimper, Air grinder, Air drill, Air shears, Air nibblers
iii.) Adapter for Standard ratcheting socket tools, torque wrench, and measuring calipers
Design option evaluation:
e.i.) requires battery usage
e.ii.) requires access to Air-line adapter
e.iii.) requires a parametric defined hole in a compliant joint
Design resource requirements:
e.i.) physical testing for fatigue, 3D scan tool exteriors (community support), and battery management
e.ii.) 3D scan tool exteriors (community support), and Air-compressor lines
e.iii.) physical testing for fatigue
Design tooling requirements:
e.i.) 3D scan tool, 3D printer, hand tools
e.ii.) 3D scan tool, 3D printer, hand tools
e.iii.) 3D printer, hand tools
Design choice selection:
e.iii.) minimal barriers to building as described in d.
i.) build prototype out of cardboard, scissors and hot glue
ii.) given the goal of c., one must evaluate with b. with g. in mind
iii.) build prototype given is satisfied
Verify design with users by checking if a. is satisfied
k.) given c. , verify b. works, and contact users in the field to see if a. is Finished
The common mistake for the author was not starting by building a single finger to test before diving into CAD.
…Consider 5 miniature robotic arms as a complexity scope estimate.
You did a great job. I find the first of any build is the hardest and longest to do.
I think that you did a great job.
Now you have a good idea of what the problems are and you can work from there.
Like Salec said start off with a skeleton frame first so that you can get at the inside to fix things.
Keep up the good work And thanks for the post.
At 4:41 – The 3d printer settings clearly need to be improved. It looks like it is over extruding and globby for the white and under extruding the black on the fingertips. With good settings there should be no globs and all edges should be within 0.1mm or so from the design. I print small (10x10x2mm or whatever) boxes until they are great and then try something larger. To get the best prints the filament should also be stored in a sealed box with known low humidity (<10% relative humidity or better) since the water expands when it boils in the hot end and causes extrusion when filament isn't being pushed into the hot end.
Working with injection molding, I have noticed the same effects. Moisture in the barrel can cause enough air pressure to build up to sound like a gunshot when purging.
3d printing can be amazing, but it has it’s downsides as you saw. Given the shapes you’ve got there are a few other ways to quickly achieve the components you want, which may be more robust.
First I’d consider deconstructing your current model, giving the old pieces a good layer of bondo or something sandable, and then making silicone molds. Then you can do castings in any material you want.
https://www.youtube.com/watch?v=i_NMFz5ards You might consider overmolding the grip pads as well. You also have the last option of casting from these molds in wax allowing you to iterate on the design in physical space, which you can then make new silicone molds from.
Second I’d consider carving. Whittling may sound old fashioned, but you’d be amazed how fast you can shape something out of a block of cheap poplar. Even without ‘proper power tools’ you could shape many of these components in a few minutes, probably faster than the 3d printer can make it. If you don’t need the strength and just want to iterate on a shape you could make some of these from insulating foam in seconds.
Overall I really like what I saw there, and I agree most strongly with expecting the first piece to have issues. I’m not sure 3d printing is wrong for your project, but it might be worth considering some alternatives.
Not sure of the plastic used. I would suggest that if it is PLA, do a lot of sanding to smooth the material surfaces. For ABS, do the acetone vapor smoothing trick. Then sanding. Smoother surfaces will reduce the friction significantly so that things operate more smoothly and have less chance of breaking things.
This doesn’t look like a big failure to me. With stronger servos and less powerful springs you might be able to prevent the servos from breaking. Also, instead of hooking the servos to thr arduino 5V outout, run them off of your 5V power source directly. I hope you can fix it in the next iteration because that arm looks AWESOME.
I think he has forgotten that he is attempting something that engineering teams with unlimited budgets would take years to accomplish. He should be proud of his progress so far instead of giving up at the first prototype.
He’s got a good idea. I hope he keeps going and manages to work out some of the bugs. There are many people around the world that could benefit from this project.
Great 1st iteration, but as stated before, start with a finger and expand.
Convert the servo’s to I2c (Attiny85?!? for minimal cost), then you wil only have 4 wires coming from the hand.(power, gnd, data, clock)
You could then have different i2c adresses for reading pots and others for turning the motors.
When you look at a smart phone it’s simple. Hidden from you are the many many hours and experience it took to make the plastic molds , the electronic circuits, the software and the the technology integration. Also hidden are the multiple revision levels required to get the design right, as well as the many people that make it happen. Your project is not a fail but rather a learning experience.hopefully you will take some of the constructive comments given here and make the next version. Great job.
With all do respect, this is not an argue, but when one looks at a smart phone, there is everything but simple. In fact, a smart phone is the most complex man-made object.
No other object ever created has a higher density of mankind achievements than a smart phone.
Billions of man-hours of research, discoveries, inventions, engineering and hard-work, all squeezed in a single object that can fit in a pocket. Radio waves, electricity, nanotechnology, digital communications, Internet, information technology, math, mechanics, mining, optics, materials, logistic, marketing, automated manufacturing, design, and the list can go on and on.
Then, it’s all the infrastructure that makes a smart phone connected.
The power grid, radio towers, communication satellites, GPS satellites, transoceanic cables, Internet, server farms, and ultimately the humans from the entire planet.
If anybody knows a more complex object then a smart phone, please name it!
Impressive design work- the shell alone had value, but the mechanics need reworked.
Impressive attempt, and a he mechanical internal redesign is totally doable.
New servos with more strength are definitely called for, and there perhaps could be more compact ones.
Basically your talking more money unless something like air muscles or twisted fillimants are used.
I’d practice on a hotdog first
Round of applause for persevering and good for sharing. Too many people think products pop fully formed by magic. For learning purposes I’d say building smaller test articles to confirm the quality of the prints and to debug mechanical problems is one area that often avoids problems, even in much more involved projects. The other observation is that while the packaging is something that would be desirable for replacing a hand, ignoring that evolution put the muscles well away from the hand should be cause for some more consideration if an arm is being developed. Still, it was an ingenious method to re-work the servos into the available space and the imagination to see how to fit those parts together is useful.
All in all, a great effort and a worthy learning experience for the creator and, hopefully, anyone else who undertakes a project of more than trivial complexity.
I like where this is going with the encoder in the hinged part of the fingers, though instead of cutting up more servos, you might check out some of the micro gearmotors on sparkfun or servocity. They’re ridiculously tiny and put out a great amount of torque with metal gears.
You’ve inspired me to post something about my failed project!
Please be kind and respectful to help make the comments section excellent. (Comment Policy)