@@jklee8866 I watched that original video a while ago and didnt expect anybody to actually attempting to build one. That is so cool, can't wait to see it working!!
Totally agree. But the "fascination" seems to distract people from the severe conceptual flaws, resulting in flex, vibration etc. Also, this concept isn't tolerant against elastic deformation, so wobbly 3D-printed arms will make things worse. Lee needs super-light but super-rigid composite arms with maximum resistance against torque momentum linked with large-diameter bearings that have zero play.
Would a 6 axis mono arm work the same way with a smaller foot print? This is very cool tho never would have thought you could make 3d Motion from 1d movement
swap the bolts for shoulder bolts add a thrust bearing at each joint remove backlash and those arms are seeing a lot of torsion loads maybe 3~5x the end effect weight, you might want to try polymaker COPA this stuff is hard as nails compered to pla or abs the bounce is shifting backlash + arm flexibility
JK Lee: "Hey guys, my 3DOF linear robot system has some skipping and bouncing, any ideas how to fix this?" Me: "Hee hoo stepper motor make pretty music"
Great implementation. Torsional stifness of each of the arms, pivots joints and linear bearings is what's holding this mechanism. The ideal shape of each arm is therefore a cylinder in first aproximation, a truncated cone with the tip toward the end effector in a second approximation when considering the "length to fulcrum" issue. The length of the pivot should ideally be longer to avoid dislocating the bearing in the worse scenario. Using a fork-like assembly (like on the GoPro mount) would be ideal. The linear bearing does not need to be long in this case, rather they should be wide. This can be obtained by coupling two bearings runing in tandem (in parallel). Be careful however, if the tandem is significantly wider than it is long it will tend to binding. Properly implementing those solutions will greatly improve the smoothness of the apparatus.
The linear rail issue is not that important IMO. The machine is very asymmetric, it is gravity preloaded. And another, cheaper way to mitigate it to lower the pivot point.
I agree with the two linear bearing solution. Use 2 rails with 2 linear bearings will improove your stability. Then you have to improove your link, they have to be as stiff as possible without friction. Also, in your case, mass is really inportant to keep the system stable ! So make it as light as possible. You can maybe use carbon tubes with 3D printed links at the extermity. Otherwise, really nice and creative design !
I think you need to stabilize the joints between the arm parts. Standart roller bearings always have some runout so they can move wich makes your arms move outside their respective plane. It would probably help to use 2 bearings on each joint to minimize unwanted motion. also the arms made of plastic are meybe too springy. you could print them with a hole lengthwise through them and glue in an aluminium rod to make them stiffer. Good luck on improving your concept it´s really a cool printer!
@@m3chanist I think the intent is to place both bearings on the same side of the joint. The pin is fixed on one side (no play) and better supported on the other.
@@dammitcoetzee These are plain bearings, if you preload them or not will make no difference. 2x bearings = 2x bearing play. Preloading will reduce the play in each but simply preloading one will still be twice as good as two equally preloaded.
Yeah looks like binding to me too, try some lubricant on the joints to see if there is any improvement, it's the quickest and simplest way. If that doesn't work you may have to rethink the type of bearings you are using. Maybe polishing any rubbing surfaces
Disney Research - "Vibration-Minimizing Motion Retargeting for Robotic Characters" could be a useful method for dampening the oscillations by adjusting the motor profiles.
thats not the problem here it might be the geometry of the joints. the Disney minimization is an optimization for vibrations caused by irratic movement
@@o_o-fk7ym the disney paper is mostly for minimization of oscillation for non-rigid mechanisms caused by movement - this monorail starts bouncing because of the movement on a non-exactly-linear rail, which I think the paper should work on
@@aonodensetsu Yes. The method should apply to any system which can induce an oscillation due to it's own motor inputs. Play and slack in the beams and joints is the same as the problem they tackle, even though the beams here are designed to be significantly more rigid to start with.
Low stiffness due to the arms was my suspicion for this peculiar monorail design right from the start. Especially for high z-values where the head-attached links come closer to lying in a common average plane (their force vectors get less linearly independent) and stiffness normal to that avareage plane suffers especially badly.
To make informed decisions: First find out about the amount, anisotropy and (eventually nonlinearity) of the compliance for forces acting on the head. That for some key locations like e.g. low z and high z. That means making experimental force deflection curves for plus minus x y z. Could be done with a string and a increasingly filled water bottle and turnig the robot every which way by hanging it up on ropes different ways.
As for the results I suspect that for high z values the compliance (in mm/N) in z direction is especially high and bad. In that case maybe the middle arm could be mirrored/flipped upside down somehow maybe??
Thanks for your detail comment.! I also agree with the point your mention - head touch the link and incread restance of the rotation! I will solve it first. Thanks.
Have someone plasmacut the arms from steel. And the first joint needs to be steel as well. Also you need to dampen it. Can you put small springs in the acute angle of each joint. That could worsen the harmonics but it could dampen it.
Maybe use the printed ones to cast aluminium ones. Also, since all the weight is on that side, I suggest a counterweight in the form of an extra rail and linear guides on the back side.
The loading on the rail is very one-sided. This is arguably an advantage because it pre-loads the linear bearings. However it seems suboptimal. For each of the outer two (the two vertical vertical-plane) arms you could put on a counterweight that hangs over to the other side of the rail. This would reduce or remove the sideways bias in the loading of the rail. A weight could be added onto both inner and outer sections of the two arms: It would make the rail-side sections into triangles and the inner sections into longer members that have the weight protruding way back away from the camera, over the rail. Stiffening it all up as other people have said is crucial. Spaced bearings. Tube sections. It is possible that the counter-weighting thing could also be applied also to the middle arm. Not sure.
A few possibilities, 1 reduce the mass on the end of the levers, 2 add dampening to the levers, 3 add stiffness to the levers, 4 reduce the weight of the levers. In 3d printing there is often a desire to extrude everything you make through a 4mm nozzle. But this is rarely the optimal design. Your overall design is superb, but it looks like a good time to optimize component materials and design. The potential is huge. Congratulations.
I suspect stiction in the joints. The stick-slip is causing the jumping. The root cause is high cross moments on the bearings. Properly addressing the loads at the bearings will fix it. For example. Dublex back-to-back angular contact bearings at each joint. Another good bearing type is crossed-roller bearings. Either option handlea the combination of cross-moment and radial and thrust loading
Also the moments can cause the gap to close between the arms at the bearing joint so the plastic arms are rubbing and giving that friction I was talking about. This could be from loose fit to the OD of the bearing. So you need the right bearings and the right fits
Thans for your comment. I checked for your point. There is no arm touch or friction. I had used two bearing as 1 set on each arm joint with preload for standing against the moment load. Finally, I have found that Z arm bearing bore deformed in moment direction and have separate movement while it's move. It's removed from new printed parts. The movement much better than before. But, Not eleminate all. I'm checking on other arms. Thanks.!
Nice :thumbsup: My guess is low stiffnes in the joints. Using wider joints with double bearings to support them would reduce the moment load of the bearing and any "wiggle" they might have because of it.
I would recommend the following... 1). Consider re-working this so that everything hangs down from an overhead monorail 2). have each lever extend out over the monorail the other way and have a counter balance.
My first attempt to correct the bouncing would be a simple procedure. Place a secure point at the top dead center of the extruder (eyelet swivel). Tension string between the eyelet swivel and an overhead pulley. At the other end of the string a counterweight located in a remote area equal to the extruder mass. It’s a beautiful machine you created. Best of luck to you.
I've never seen anything like this before! Cool! If I could guess it's either A: the bearings on the printing head aren't sufficient or there's slop in the joint B: there's something creating a perpendicular movement in the printer head joints which are causing the axel to buckle thus causing the bounce.
@JKLee Try reducing the backdrive on the middle arm! If this arm could provide a little more control it might help stabilize the changing angular torque values applied by the outside arms! The shaking occurs when these two arms move, it seems! That geared arm you just showed off, should help!
try using shoulder bolts and properly toleranced bearings for the joints, make sure the bearings have a moderate press fit into the printed parts. a normal threaded bolt won't have a precise enough OD to fit perfectly into a bearing like this.
A single axis of movement along the extrusion makes sense, but the head end should be ball joints to support full 3 axis of movement, on all three points. You are binding (bouncing) because the head end is restricted to a single axis, when really each point at the head end needs 2 axis of movement, the easiest solution would be to use ball joints on the head end which would allow 3axis per arm, where only two would be required, but with 3 you would not have to align/clock the two Axis per arm. The middle arm controlling y/z retraction can stay as a single axis movement. The outers need to be ball joints on the head. This is really cool!
as some suggested stiffening up your structure is probably first thing I would do specially your adapters on the carriages, they should be pretty stiff. If alu is not an option, you can cover them with fiber and epoxy, like they do on boats :). Stepper motors have a lot of ripple effect, on a system that is not ready for it, it might multiply largely to cause all the bouncing one solution can be to use BLDC motors, but they can be pretty expensive or sort of a smoothstepper.
Honestly even with 100% infill and high quality bearings in all joints, this will still not be a stiff design.. but low stiffness machines can be compensated with advanced motion control schemes. But thats a really deep topic..
Good point, but you shouldnt really care about these points in early Prototypes. The levers could be made out of aluminium or fiber reinforced stuff later on
Once you have implemented the other ideas, to reduce the effect of slack or binding in the joints whilst all the joint are under twisting force as far as dimensional and budget limitations allow, you could look at secondary considerations like a. moving the COG of the head assembly directly under and as close to the intersection point of the three planes as you can to reduce the effect of the wobble; b. detuning any resonance of the system by adding tuned dampeners like buildings use (especially when there is one significant resonant frequency) or using something similar to how a dead blow hammer spreads the force from movement eg build the arms with internal spherical voids and 1/4 fill with sand part way through the 3D printing and then complete the 3D printing to enclose the voids; [I thought of suggesting making the outer arms lighter than the inner and thus resonate at different frequencies, but you don't want to induce an action like a whip]; c. ramping up and down the stepping frequency and/or the phase of stepping so the effect of the each step is out of phase with the last and thus dampening the bouncing that is due to resonance; d. and lastly, after you have done everything else, if some of the bouncing is due to joints slipping once an initial resistance is over come, deliberately inducing small jiggling/resonance in the stepping, even to the point of sometimes stepping in the wrong direction, to get or keep the joint slipping rather than stopping and binding may help. I hope my suggestions help, from a software professional who enjoys thinking about hardware, but leaves it to others to spend the money actually making stuff ;)
Thanks for your well summarized comment! I will try to implement head design - good to know COG under joint to reduce effect. Additionally, I will also consider stepping frequency due to Z axis and X/Y has different step rate. Yes, there is many things are not optimized. But, your comments will help improvement. Thanks!
This is actually amazing. The simplicity of having one rail and the massive scaleability means this could be a very promising design for large-scale industrial FDM 3d printing. amazing work!
I would guess this is a rigidity issue, if there is any flex in the arms or joints it will be used instead of the wanted motion, if the force required for the flexing exceeds the force required for the extruder assembly to move then the bounce will hapen. To fix this you could try to counterbalance the arms so that the effective weight being moved is as low as possible and is always less than the stress that causes tortion. Counterbalancing can be done by making each arm extend over it’s pivot point and adding weights to the back. Completely removing the bound will be a balancing act between rigidity of the members and weight. Lowering the acceleration should also help
If you take the belts off, a correctly balanced machine should stay in whichever position you place it in. Getting as close to that as possible should be the goal for this machine. Double bracing the joints might make them more rigid to counteract the flexing too. Try replacing the members with 2020 extrusion for this extra rigidity, then double brave each joint
@@AxfordIndustries thanks for comment. I tried to adjust tghting of bolts and speed/acceleration decrease. There is only changes its bouncing frequency.
I think the arms are too springy with plastic. Might need to try it out with carbon fiber rods or similar instead to increase stiffness. Also as others have said you should probably have thrust bearings, at least for the middle leaning arm.
Before you can really smooth things out, you need thrust bearings with preload. The bearings you're using are suited for radial loads, but you're dealing with axial loads. Any slop in the bearings, which radial bearings have by design, is amplified in the print head. It may not fix it entirely, but looking at how smooth the linear rail bearings run, it doesn't seem like the controller, arm stiffness, timing belts or steppers are the issue. When in a fixed position, I suspect the print head has a lot of play in it. Thrust bearings with some preload can reduce this drastically. You need at least two pairs per joint though, and the axles for each joint should have no play. Precision shoulder bolts sized to the bearing would work, but they can add weight. Accurately sized pins with c-clips can reduce weight, but you need to size them properly to get the right preload. Others have also suggested flextures, which is a great idea too. Using spring steel instead of pivots has a lot of advantages (like having your arm segments share the same plane). They can also give the linkages some preload. It can last too - the plastic arms will wear out before the spring steel if design right. Look up Dan Gelbart on youtube, and find his videos on flextures.
Thanks for your comment. I also agree the possibility from controlling or stepper timing point. So, I'm considering to modify that point. Anyway, many people pointed out the two pair of bearing and preload. I will prepare and share what the arm joint design.
You can do modifications of different tiers: - tune the stepper profiles, implement active damping (low cost, just time) - use stiffer materials for your design: 3D-printing dynamic load bearing parts is asking for trouble, i.e. replace the orange arms and the mounting bases (more cost, potentially outsourcing the manufacturing of the parts) - scrap that design and opt for some kind of delta configuration (pragmatic for stiffness, but not as flashy). As another comment mentioned, there are inherent flaws with that design, most importantly that you can only get so much torsional stiffness out of a MONORAIL. Very cool concept though, and I'm convinced you could get to something decent with the first two solutions
We had similar issues with the original Morgan. The flexibility of the arms is too high, even if it is very small. To solve this we eventually replaced the arms with large diameter carbon fibre tube with the bearing mounts printed. With this machine, even the smallest flex in the arms will cause a jump. Replacing your arms might need something even stiffer like aluminium alloy. Fascinating movement though.
Correct me if I'm wrong, but it looks like these joints all work with limited range of motion, right? When I see limited range of motion + stiffness requirement, I think flexures. What if you could turn those bearing pivots into flexural joints? Easy enough to print, the design itself seems like the hard part. I've seen the double butterfly joint used before, though I'm not sure if it's the best for your situation. Just an idea! Cool stuff, keep it up.
Do flexures offer repeatability? And don't they break after a certain amount of movements? I always find those fascinating, but they gotta break after some use right?
@@MrRonny6 I don't know enough about fatigue failure in plastics, but yes, any material will fail due to fatigue eventually if the loads are sufficient. For this application, the loads are relatively low and the number of cycles to failure would be high enough to prove the concept.
Pre stressed ball bearings. Flat / axial bearings. Some bounce will be unavoidable given how long the rods are - all the slop, all the flex will be multiplied by the length... Still an amazing concept. Minimalistic, good for "general purpose" positioning... will be hard to make it precise though. Will require extreme rigidity and stiffness - or additional sensors and complex feedback loops.
The explains would be interesting to see. Without clearance / restressing the ball bearings, it will have some wiggle in the joints. And friction will cause some "skipping", "jumps" - bumpy movement. Another thing to consider is mass. Adding extra mass to the cargo will slow the thing down but - change the natural frequency of the structure too. The wobble can be made "slower" if the mass is increased. And if the cargo mass is reduced - the cargo will oscillate at a higher frequency. The stiffer the rods and lighter the cargo - the higher the natural frequency of the structure. And using steppers - some oscillation will not be avoidable. Servo motors would be more smooth.
Everybody talks about making joints stiff and preventing binding and I'm wondering do we have enough resolution on the belt driven axes. We basically translate rotational motion (stepper) to linear motion (linear rail carriage) to rotational motion (arm) again. Any positional error of the stepper motor due to motor imperfections or microstepping is basically multiplied by arm length. From what I see the two outside arms (belt driven) should never move independently - one arm stationary when the other moves as it would bind the whole mechanism. As the arms must move precisely in sync so must the steppers. Any stepper positional error is translated into force stored into the belt (which acts like a spring) and released when positions get in "sync" again which is visible as wobbling, especially when doing translations on the axis perpendicular to the rail (only outside arms move). From what I gather there are three ways to fix this: 1. Make arms driving mechanism stiffer (switch from belts to lead screws) so two steppers would basically force each other into position "in between the error" - this could lead to real binding and damage to the arms and/or head. 2. Make arms and/or head more compliant so the stepper positional error won't be stored as energy in the belts - this could probably lead to inaccuracies in the head positioning. 3. Put more positional accuracy into the system by using more precise stepper motors (0,9 degree), gear reduction (backlash!) or belt reduction (smaller pulley?) - this is probably the best compromise. Anyway great job so far. I hope you have enough perseverance and helping hands to actually polish this project as it would lead to seriously cool and portable 3d printer.
The applications of potentially using this one day to simplify the printing of foundations to houses by potentially having something like this mounted on a long trailer, pulling up to a lot, and just “printing” with concrete is really cool to think about. This tech will be bigger than we can imagine one day.
Looks like most people have covered this already but the further the point you are trying to control is from the source of the movement the greater the amplification of any small variations or gaps. The arms are essentially fulcrums and any play in the material as well as the joints is going to cause what is seen here. You would need some vary rigid materials and some joins with very very tight tolerances to make this usable.
Looks really cool, and a its nice to see it proved from the renderings, but it seems that any tolerances are magnified a lot, belts, bearings and arm flexure ect, when you get it printing make sure you try a large circular shape and see how well it can manage that.
Fascinating gizmo, I could see this becoming a really long 3d printer when you get it tightened up. Not quite a belt printer, without the issues of a belt printer. Bonus points for it sounding like a 1980s arcade. Makes me want to pop a quarter in the Asteroids machine.
This is by far one of the most fascinating thing I've seen moving. Looks like @Kelvin Nishikawa has the best idea for fixing this with the thrust bearings. I also thing maybe if the arms were something more rigid, like cut aluminum, though that would be a challenge to acquire.
Remember that a robotic design with some flexible "springiness" inherent in its construction can be safer in close proximity to humans; "softer"(a virtue) in that it is less likely to have high-G-rate(damaging) crashes. Depending upon how it is used and what applications it is used for, this might be a virtue not a vice ==>Interesting mechanism, may have many good applications, food service first came to my mind. Joint free play control, and how it affects/effects overall unusual motions would be where I would usually start(but this mechanism is new to me) Kudos on your work, (subscribed✔️). As I read more about the kinematics and your intended function, I can be more helpful for solving your problem,,, just to remind you that correctly used(adjusted) the fact that there is"bounce" may be a good thing (tuning it down to the correct amount for your application, is the question) Again, interesting stuff, well done👍
My recommendation would be to put all arm mounting points in double sheer. I don’t think the rigidity of the plastic is causing your oscillating bouncing. I think when the motion changes direction it moves the load on the pivots from one direction to the other and the pivots are flexing slightly because you’ve turned all pivots into a moment arm. Having them all in double sheer I think would improve the machine.
It looks like static friction is the problem. As the arms move they are building forces but not enough to move, and when enough force is applied to overcome static friction it snaps, like an earthquake. I would recommend disconnecting the motors, and movie it by hand to see if this is the case - follow a similar path and see if that is the case. Normal bearings can cause this effect too because of slim sleeves able to have too much pressure applied and end up deforming them slightly for cheap bearings or bearings using different material types. You'll be able to feel where the problem is - our hands are incredibly sensitive, so when a snap happens, just try to follow that - or have someone else movie it and you touch the arms, etc.. and feel where those jumps are. Edit: And bearings do have gaps - so it may move a little and then impact requiring move force to move or rest on the wall already.
The attachment of the middle arm is causing some of that by binding against the motion of the other two arms; the other person who mentions thrust bearings is right, but you also need some sort of counterweight, the weight of the head is more than that of the arms and they're not rigid. If you don't want to redevelop the center arm so it's not binding as much against the motion of the other two, I'd look at some sort of counterbalance
And speaking of this in terms of robotic arms and stuff, usually there's springs, counterweights, etc, to reduce that mass shifting or flex in the armature.
What I'm intrigued with your implementation there is how you're driving the arms with the single rail. Interesting choice for a limit switch, by the by. It's a bit of a deviation from the typical microswitches or optical solutions.
As has been noted, the lower arms are mostly loaded in torsion. Box the arms (ie. add multiple top layers) rather than the single face or try manufactured tubes with printed sockets.
This is an awesome and fun proof of concept! To be honest the level of bouncing you are getting is pretty good considering how minimally constrained this is - the monorail setup is placing the maximum possible demand on every single joint, bearing and plastic part. I believe the main source of the bouncing is normal 'ringing' from the belts, but able to resonate through the many joints fighting one another. Tighten the belts a bit more but not ridiculously tight. Thrust bearings could help. Lighten the arms further - carbon fibre square tube or thinner CF strand reinforced plastic would be good. Put dampers on the stepper motors, and this might sound a little mad but a little vial of liquid on each of the moving carriages might help absorb some vibrations. Not sure what's reasonable to expect quality wise from a 3D print from this. but I hope you can get it to functional!
your toolhead is bouncing due to binding, either in the bearings themselves due to off-axis loading, or the segments of the links rubbing against each other. You could try using a small amount of a PTFE bearing silicone lubricant between the segments where they are rubbing and see if it decrease the binding. Another thing to try is doubling the amount of bearings between each segment, which should help reduce the side loading on the bearings. The other consideration is the segments in the arms themselves twisting, since you are relying on the arms being rigid members. Ultimately, while I like the idea of this kinematic system, I feel that it likely has a lot more bugs to work out than the bouncing you are getting at low feedrates. Good luck!
a couple of ideas that come to my mind : - i believe you move the axis in curve (non-linear steps) maybe you can refine your curve more with even finer steps? - use higher bit microcontroller, eg. used 32 bits instead of 8 or 16 bits - make use of ball joints instead of "elbow" joint
Hello, Have you got a useful answer? Here is a list of possible causes of this bouncing: 1 - Rigidity of the material of the arms. 2 - The rigidity of the support of rail on the table. 3 - The play in all joints of the arms, and how the ball bearings do not play in their holes. 4 - The precision of the guide on the rail. 5 - The elasticity of the belts. 6 - The precision of the screw and the play of the nut. 7 - The way the screw is maintained along its axis and the coupler. You cannot rely of the motor rigidity nor on a soft coupler even though you may need one. I do not see o the videos how the screw is mounted at each end. This kind of nuts is a sacrificial part of 3D printers that force only in one direction on the printers. Some printers incorporate a system composed of a spring and a second nut that constrain the main not to always be in contact with screw. But this works is the nut is acting always in the same direction. The first thing you could do is to make static measures. Block the motors in one position where the head is 100 mm away from the rail for example and maintain the motors (do not disable the drivers). With a comparator you will put a bit everywhere to measure the displacement while applying manually a force, and try to evaluate - the play on your system (slightly pulling up and down by hand, feeling when the system is constrained in one direction and the other) - the stability of the thing, by forcing a bit more. Regards
Accumulated play thats perpendicular to the axis of rotation of each elbow bearing, twist. You need bearings with some preload rather simple pivots, small angular contact or taper rollers ideally to handle both load directions. The arms need to be very stiff in the bending moment, not simple beams, plastic is not ideal.
The bounce is due to the arms negotiating stray torque forces. I assume the arms always maintain a air gap and are linked only by the inner race of the bearings minimizing the chance of stiction (stiction would cause a similar issue but unlikely at this stage of the project). When you have a device with two arms they tend to self balance the minute differences in stray torque with one taking the lead in moving and the other following, but with 3 or more arms they will bounce due to force being loaded on one arm until it bounces free and then force loads on another. If you had a perfect mechanism with Idea servo's and magic math you wouldn't get stray torque but in real application its impracticable to implement a level of accuracy to eliminate stray torque so the best bet is to accommodate it. To reduce it effects you could introduce a ball joint on the rail end of two of the arms the arms or the outer two cradle ends of the arms.
Add a rubber band between the end point and the arm connection on the rail. This will keep a more constant pressure against the bearings and should smooth it out. Like shock absorbers do.
There's a reason tripteron is not a commonly used design, and that's also why SCARAs look surprisingly beefy. Joint on joint on joint can be a source of flexing. This machine needs more stiffness. My suggestions: First, use much larger bearings (3x dia. or more) and preload them. It will increase resistance, but it's a trade-off. You don't need thrust bearings. Large, thin deep groove type is probably fine. I'm sure the arms are twisting. Since it's mostly the perimeter that takes this kind of load, you can make them larger but hollow to keep weight low. Thermoplastic alone is not the best material in this case. Try to reinforce them with a fiber composite shell. Tripteron kinematics is mesmerizing and your monorail version is a beautiful take on it. The practical implementation will be a challenge though.
What I see is that when your two forward arms are extended too far in the x direction along the track, you are getting a small amount of twist though those same arms. When the torsional pressure is released from the extension of the arms is when the bounce seems to occur. Y-axis rotation at the base of the two forward arms may help solve that.
I think the weakness of this system is that it relies on the arms and bearings taking forces that are out of their movement plane. This causes twisting torque on the joints that increase friction, binding (alternating between rest and dynamic friction) and potentially slack. You would need more robust multi row bearings and joints that can handle this well.
After watching this mesmerizing video for a minute or so... Looks like compliance (material is bending) in the 3D printed arms. I think I noticed this mentioned in some other comments. Look up Delrin, its a fairly stiff and very machinable plastic that could work well for you.
it's the pulses in your signals, you hear as different music notes, being amplified by the long arms if the signals to the steppers motors changes frequencies more smoothly, then the bouncing would go away it's a combo of the harmonies between the different stepper motor speed changes
MY humble idea on your bouncing prob. : Arms having intermittent frictions issues with each other at the rotating joints because there is sideways forces pushing them at each other. But wachers to separate them.
To fix up and down oscillations, The center arm needs a second arm operating semi-symmetrically on the same rail. Kind of like a really complicated scissor-jack. Basically, that center arm, needs a triangle with 2 points on the same rail.
I think it has to do with the bearings in the middle arm being twisted into angles it doesn't like, giving that weird effect. Might need higher quality/bigger bearings, maybe even a different style After more thought adding a bearing to each joint, it would change the center of rotation to the walls of the bearing giving it a more 'normal' load instead of twisting them.
The problem resides in the stiffness of the joints. They need to be extremely rigid and precise with no backlash. They also need to be fluent and friction free though. Another problem might be the rigidity of the arms, if they are too flexible the bouncing effect is amplified.
Possibly longer linear bearings and associated plastic mounts. I suspect the twisting forces here are the largest contributor to loss in rigidity. Cool machine.
There are many solutions which you probably already plan to implement on the next iteration, but just in case, I will describe them: Oversized the cross section of the arms and print them with a stiffer Carbon Fiber fill filament. A professional version of this would ideally have cast Aluminum arms for rigidity. There is just too much cantilevering going on and every movement has a vertical component to the forces exerted. This is the nature of the thing with those 45° mounted bearings. Every effort must be made to make this thing as lightweight as possible. Lowering inertial mass will reduce the bouncing. With this much undesirable slop/motion as shown in the video every technique possible should be implemented. For instance: Reduce the weight especially out at the ends of the arms. What's the lightest weight Hotend you can find? Also, you could upgrade the bearings to cone thrust bearings but this will be expensive and difficult to source the parts. I believe you can get acceptable performance from the 608 Skateboard bearings but you will need to space them with a larger gap for more rigidity. Each end of each arm should have a pair of bearings mounted maybe 30mm apart. For a total of 4 bearings per joint. Your design will need a way to adjust the axial preload force on the bearings. This will eliminate slop/backlash in the joints. Good luck and keep up the fantastic work. I love this project!
@@jklee8866 Speaking of lightness and rigidity have you considered making an arm from two printed parts, at each end for the bearings and one middle part using one of those standard premade carbon fibre rods
The monorail design puts a lot of strain on all joints and bearings. I believe that was obvious, when you designed the mechanism. So you need to address this as best as possible. A first and probably easy step would be to double the orange arms attached to the holders on the rails. This might even be done without any design change. At least for the left and right arms. Just lake longer screws, and add an opposing arm with bearings. That should increase the rigidity drastically. Looks like it might fit. Maybe even also double the secondary arms connected to the head. That would mean 4 bearing on one axis and way more stability.
Thanks for your comments. Especially, the arm of Z axis is week arm to bear the load by head. I will consider two arm or wider arm as your comment. Thanks!
Just tossing this out there in case it's helpful, I do think the others are right about binding, but if that *doesn't* solve the problem, I would look into the controller. This thing has to do some pretty weird trig/IK to put the effector where you need it, and I know in delta printers you can get issues if your controller is a bit slow. It's not likely the cause, but it is something to look into if the problem sticks around.
The problem is in the geometrical design of the machine, there are long plastic pieces that hold the extruder that have some flexibility, also the bearings allow a little movement, and the geometry of the machine makes all the play and flexibility to multiply, a more conservative design with screw rods in the xyz axis will have less vibration.
@JK Lee: this could be because the balls and the running surfaces in the ball bearings are not round enough. The ball bearings are available in different precision classes. The more precise the more expensive - of course.
It's probably the actuators resistance to each other. As an example when the leftmost one is going right the middle one has to fight that force which is being applied to it by the fact that they are connected all together. That produces bounceback movements and overshoots that are not accounted for by the control system. I have no clue on how It works. Maybe if you are using a PID controller try to raise the I term. But then, i might be completely off. To test this maybe try to hard lock and fix mount two of the actuators at a time and just move one and see for each of them if you still have bouncing problems, once you get rid of them by isolating the single movement try to account for the forces applied to and by the other actuators
Try using aluminum rectangular tubes and axial bearings combined with radial ones for the linkages. You should be able to reduce the vibrations for a bit.
Hello, my guess to improve this: Assembly is not steady enought due to (1) too much axis and (2) weak motor power to fight gravity. -maybe try with elbow type joints like crane (but not wired, with hardware materials) and grease on elbows Or add motors on joints to help, but this will increase weight... Anyway you have althought already a beautiful machine!!
seems hothead looks to be heavy for arms size.. tip end placement increases leveraged weight too.. so, lighter massed hot unit or stronger wider arms might distribute end weight easier to carry hold steadier? wider contact (perhaps polished steel washers few cm's larger) by bearing rotors joints would align movement by surface contact glide opposed to lifting strength alone.. amazing compact design concept for dimensional movement.. nice work! keep at it..
I would like to point out that the arms aren't perfectly designed nor is the material strong enough to compensate so it might be the elasticity of the material just shaking the extruder. this issue is compounded by the rough movement of the steppers, I recommend trynamic to at least have a smooth movement, well lubricated rail, joints and nut and screw. If the issue persists try changing the arms with some carbon fibre or steel rods, perhaps you could use pipes instead of rods or multiple rods in parallel to form a single arm.
I think that this solution where the head is placed at the end of something equivalent to a rod will always have some "flex" in it and vibrations will magnify. It also will have a natural oscillation frequency. As others mention to stiff things up will help.
While I don’t operate 3D printers, it appears that the bouncing is coming from a bending moment between the mid and head bearings. I would recommend that you thicken the mating area by the head and implement a long set of needle bearings. Thrust bearings also seems like a good place to try first though.
The circle motion sounds like music. I feel like the motor holds a certain speed for a short while and a moment later it changes it speed with a little bit instantly. It doesn't sound like a gradual change of speed. Maybe these constant shocks through the arms result in a bit of bouncing too. Of course this is just a hypothesis and it could be wrong, but if this is not the issue (or one of the issues) then at least you can be more sure of the right answer :)
Suggest you look at the Tripteron design done by Aspu there. The monorail idea is intriguing, but the basic simple joints you're using there lack sufficient stiffness along with the printed arms (Aspu started out with 3D printed arms and then moved to printed hinges, etc. and 2020 rails, etc. ) which is where at least part of where the bouncing is coming from.
1: Conical ball bearings so you adjust away ALL play 2: Extremely stiff arm-parts. Cool idea! But everything needs to super-mega-stiff and totally over dimensioned for the load.
I'm not an export .. But you may check "basics" points : - dose all arms move freely ? - did you see friction on each parts ? -> in fact the motion look jerky only on one direction => did you try linear motion 1st ? (only X / Y / Z and check if you are able to have linear motion - check "speed" too to see if you have speed variation -> set only one axes and record the motion then check frame by frame ... the other method is to draw the line and move the paper in the other way -> you must have line to but if you have a curve, it's means the speed is not constant according the angle. Camera record could be easier to test ) As I can see , the orientation of each arm are not the same, so it can be a speed issue (you have 3 motors but you have 2 arm axes so one arm must move faster / slower as the others) => could you try to add a pen on the head ? I'm pretty sure you may not really circular as I can see on the lase sequence (2:20 side view) the Z change but (for me) it shouldn't -> this is a very promising prototype !!
I suggest to change two of the end effector bearings with ball joints, as the current design requires the rotation axes per link to be accurately parallel.
The first thing I would do is add z-bend to the 3 "forearms" so that the forearm facing sides of the bearings on printing head were in the same planes as bearing faces on the rail brackets (right now these planes are offset by the thickness of the forearms). Next thing is to locate the points of intersection between those planes and corresponding axles of the rail bearings. All three points should be on a same straight line parallel to the rail (right now the middle point seems to be too far back, and I am not sure if side points are aligned either). These changes should eliminate a lot of small but randomly directed forces on the bearings. If it is still not satisfactory, I'd replace ball bearings with roller bearings.
You could try a finite-jerk motion planner. Or, possibly, just use links that are stiffer. When theory and practice differ, it is frequently because you have forgotten that everything physical is a spring.
Amazing. I guess the connection to the rail might be weak and the material you are using also might add some movement to it. If is not too much trouble you could try with aluminum arms and stronger joint. Amazing mechanism though 👏👏👏
I'd recommend changing the arms to interleaving "fork and blade" connections at the pin joints so they can't put as much off-axis torsion on the bearings.
Perhaps think about some type of suspension between the arms such as some springs or wire to keep the frequencies of any movement/vibration even throughout. Shouldn't eliminate shake but could smooth movement.
Besides what others have mentioned regarding the stiffnes of the arm I think there could also be an issue with the trajectory. Basically If one motor has to travel a short distance to reach the desired position and one has to travel further, the second one is gonna be there later. This can result in undesired pathing of the end effector. even though the position in the end is correct.
You can test the mechanics by just driving one axis at higher speed. Break with a high deceleration and see how it behaves. If it bounces after that you have a problem with the stiffnes of your arms and joints. If the problem prevails even at low endeffector speeds it is most likely a problem in the trajectory generation
Very nice mechanism for a 3D printter! I think you must attack the vibration souce, try to reduce at minimum the vibrations. Apart from that you can always improve the joints quality or matemathically model your system and obtain from the model the value of the optimal dampers, test some of them and so damping the system. But from my perspective, just watching the video cant tell you more than it looks like your are using stepper motors and the jump between steps is to big and agresive and so produce the vibration that is amplified by the lenght of the arms. You can try to use shorter arms to see if the vibrattion is reduce and so valifate this idea. If this is the case... the are several options!
Seems to be due to friction in the bearings at small angles from arms 1 and 3. The arms are pushing against one another and its likely causing positioning inaccuracies due to friction lag. Perhaps consider adding an adjustable tension rod between the two arms to make them more connected and stable!
You have a lot of lateral force on the arms. If you watch the center arm closely, you can see it's slightly twisted due to the force, and the joint is being pried open. The materials aren't stiff enough and the joints aren't rigid enough, so it's bouncing
You could make the rail effectively wider by installing a second rail and set of sliders a few centimeters from the first, and attach the arms to both rails with a linear bearing on each rail. That'd give you a longer moment arm to resist the torque moving the head up and down.
to me it looked like the arm attached to the monorail wants to tilt at the pivot point but it binds try to give the arms connected to the monorail some room to tilt right to left and try to clamp the ball joint on both sides i'm no expert soo take what i write with a grain of salt but hope to see more update videos it's fascinating good luck man
Looks like the normal play in the bearings are being amplified by the long arms. Could try tapered roller bearings, much tighter tolerances due to their design. A cheap band aid fix could be just add springs at each joint to add tension and smooth out some of the bounce.
I think the geometry requires the arms to be very rigid and for the bearings to have no free play whatsoever. The bouncing seems to happen when large movements of the head are being produced for very small movements of the drivers. I think just making everything more substantial would solve the issue.
I'm also feel like you when I first watch Nicholas Seward's 'RepPap Monorail' concept. ruclips.net/video/oBBhVcemVHc/видео.html This is what I try to make it real based on his concept.
What a fascinating method of 3d movement, translating from a series of 1d movements
Thank for your comment.
I'm just inspired by Nicholas Seward Monorail concept.
ruclips.net/video/oBBhVcemVHc/видео.html
@@jklee8866 I watched that original video a while ago and didnt expect anybody to actually attempting to build one. That is so cool, can't wait to see it working!!
Totally agree. But the "fascination" seems to distract people from the severe conceptual flaws, resulting in flex, vibration etc. Also, this concept isn't tolerant against elastic deformation, so wobbly 3D-printed arms will make things worse. Lee needs super-light but super-rigid composite arms with maximum resistance against torque momentum linked with large-diameter bearings that have zero play.
@@f.d.6667 it’s a cool design if you want to, idk, sort the pencils on your desk and have the robot not take any space of the top hahahah
Would a 6 axis mono arm work the same way with a smaller foot print? This is very cool tho never would have thought you could make 3d Motion from 1d movement
Use thrust bearings in the elbows and attach the lower arm like a clamp so the bearings hold more forces perpendicular to the rotation axis.
came to add this. slop in the bearings from running on threaded bolt likely not helping.
I don't really know what this means but it sounds smart so it must be right. Great answer, Kelvin!
swap the bolts for shoulder bolts add a thrust bearing at each joint remove backlash and those arms are seeing a lot of torsion loads maybe 3~5x the end effect weight, you might want to try polymaker COPA this stuff is hard as nails compered to pla or abs the bounce is shifting backlash + arm flexibility
maybe lightweigh 60mm x 78mm x 10mm bearings
Came to say the same thing. There are a ton of lateral thrust forces expecially in the middle arm. There needs to be lateral stabilizing of some sort
JK Lee: "Hey guys, my 3DOF linear robot system has some skipping and bouncing, any ideas how to fix this?"
Me: "Hee hoo stepper motor make pretty music"
Did you clean linear housings? bcs sometimes theese housing are full of scrap and balls are messy to :)
Maybe some stepper voltage filter would help
ARE YOU BLIND? HIS DOING THE SAME THING, YOU EXPECT HIM TO HAVE THE SOLUTION? LOL
Great implementation. Torsional stifness of each of the arms, pivots joints and linear bearings is what's holding this mechanism. The ideal shape of each arm is therefore a cylinder in first aproximation, a truncated cone with the tip toward the end effector in a second approximation when considering the "length to fulcrum" issue. The length of the pivot should ideally be longer to avoid dislocating the bearing in the worse scenario. Using a fork-like assembly (like on the GoPro mount) would be ideal. The linear bearing does not need to be long in this case, rather they should be wide. This can be obtained by coupling two bearings runing in tandem (in parallel). Be careful however, if the tandem is significantly wider than it is long it will tend to binding. Properly implementing those solutions will greatly improve the smoothness of the apparatus.
The linear rail issue is not that important IMO. The machine is very asymmetric, it is gravity preloaded. And another, cheaper way to mitigate it to lower the pivot point.
I agree with the two linear bearing solution. Use 2 rails with 2 linear bearings will improove your stability. Then you have to improove your link, they have to be as stiff as possible without friction. Also, in your case, mass is really inportant to keep the system stable ! So make it as light as possible. You can maybe use carbon tubes with 3D printed links at the extermity. Otherwise, really nice and creative design !
Idk what I’m talking about but Maybe also lower (COG) on the entire arm structure might help
@@antoinetoinou49 way easier to just use a single larger linear rail (hgh20 or bigger) than two small ones.
I think you need to stabilize the joints between the arm parts. Standart roller bearings always have some runout so they can move wich makes your arms move outside their respective plane. It would probably help to use 2 bearings on each joint to minimize unwanted motion. also the arms made of plastic are meybe too springy. you could print them with a hole lengthwise through them and glue in an aluminium rod to make them stiffer. Good luck on improving your concept it´s really a cool printer!
Two bearings will have the opposite effect, they will simply double the play.
@@m3chanist I think the intent is to place both bearings on the same side of the joint. The pin is fixed on one side (no play) and better supported on the other.
I believe this is correct. The "knuckles" have too much play looking at it. Maybe wrong tho.
@@m3chanist not if you preload them against each other.
@@dammitcoetzee These are plain bearings, if you preload them or not will make no difference. 2x bearings = 2x bearing play. Preloading will reduce the play in each but simply preloading one will still be twice as good as two equally preloaded.
Yeah looks like binding to me too, try some lubricant on the joints to see if there is any improvement, it's the quickest and simplest way. If that doesn't work you may have to rethink the type of bearings you are using. Maybe polishing any rubbing surfaces
Disney Research - "Vibration-Minimizing Motion Retargeting for Robotic Characters" could be a useful method for dampening the oscillations by adjusting the motor profiles.
thats not the problem here it might be the geometry of the joints. the Disney minimization is an optimization for vibrations caused by irratic movement
@@o_o-fk7ym the disney paper is mostly for minimization of oscillation for non-rigid mechanisms caused by movement - this monorail starts bouncing because of the movement on a non-exactly-linear rail, which I think the paper should work on
@@aonodensetsu Yes. The method should apply to any system which can induce an oscillation due to it's own motor inputs. Play and slack in the beams and joints is the same as the problem they tackle, even though the beams here are designed to be significantly more rigid to start with.
Low stiffness due to the arms was my suspicion for this peculiar monorail design right from the start. Especially for high z-values where the head-attached links come closer to lying in a common average plane (their force vectors get less linearly independent) and stiffness normal to that avareage plane suffers especially badly.
To make informed decisions: First find out about the amount, anisotropy and (eventually nonlinearity) of the compliance for forces acting on the head.
That for some key locations like e.g. low z and high z. That means making experimental force deflection curves for plus minus x y z.
Could be done with a string and a increasingly filled water bottle and turnig the robot every which way by hanging it up on ropes different ways.
As for the results I suspect that for high z values the compliance (in mm/N) in z direction is especially high and bad.
In that case maybe the middle arm could be mirrored/flipped upside down somehow maybe??
Thanks for your detail comment.! I also agree with the point your mention - head touch the link and incread restance of the rotation! I will solve it first. Thanks.
A. this video is blowing up....
B. did you fix it?
Yeh, blowing up, I've noticed this too.
More rigid elements
@@xWhiteLeG3nDx my thoughts exactly
Have someone plasmacut the arms from steel. And the first joint needs to be steel as well.
Also you need to dampen it. Can you put small springs in the acute angle of each joint. That could worsen the harmonics but it could dampen it.
@@altimmons The bouncing comes from the joints
Maybe use the printed ones to cast aluminium ones. Also, since all the weight is on that side, I suggest a counterweight in the form of an extra rail and linear guides on the back side.
An extra rail defeats the purpose of the monorail
@@dennyrulos4847 I guess we can call it DualRail, but it's still working like 1 rail because it's both on one side.
The loading on the rail is very one-sided. This is arguably an advantage because it pre-loads the linear bearings. However it seems suboptimal.
For each of the outer two (the two vertical vertical-plane) arms you could put on a counterweight that hangs over to the other side of the rail. This would reduce or remove the sideways bias in the loading of the rail.
A weight could be added onto both inner and outer sections of the two arms: It would make the rail-side sections into triangles and the inner sections into longer members that have the weight protruding way back away from the camera, over the rail.
Stiffening it all up as other people have said is crucial. Spaced bearings. Tube sections.
It is possible that the counter-weighting thing could also be applied also to the middle arm. Not sure.
A few possibilities, 1 reduce the mass on the end of the levers, 2 add dampening to the levers, 3 add stiffness to the levers, 4 reduce the weight of the levers. In 3d printing there is often a desire to extrude everything you make through a 4mm nozzle. But this is rarely the optimal design. Your overall design is superb, but it looks like a good time to optimize component materials and design. The potential is huge. Congratulations.
Thanks for your detail comment. I will try implement your points!
Feel free to contact me directly, I learned a lot from a similar field that I put into practice. ksadil@gmail.com. love your work.
Hehe, it sounds like its singing flight of the bumblebee!
I suspect stiction in the joints. The stick-slip is causing the jumping. The root cause is high cross moments on the bearings. Properly addressing the loads at the bearings will fix it. For example. Dublex back-to-back angular contact bearings at each joint. Another good bearing type is crossed-roller bearings. Either option handlea the combination of cross-moment and radial and thrust loading
Also the moments can cause the gap to close between the arms at the bearing joint so the plastic arms are rubbing and giving that friction I was talking about. This could be from loose fit to the OD of the bearing. So you need the right bearings and the right fits
Thans for your comment. I checked for your point. There is no arm touch or friction. I had used two bearing as 1 set on each arm joint with preload for standing against the moment load. Finally, I have found that Z arm bearing bore deformed in moment direction and have separate movement while it's move. It's removed from new printed parts. The movement much better than before. But, Not eleminate all. I'm checking on other arms. Thanks.!
Nice :thumbsup: My guess is low stiffnes in the joints. Using wider joints with double bearings to support them would reduce the moment load of the bearing and any "wiggle" they might have because of it.
I would recommend the following... 1). Consider re-working this so that everything hangs down from an overhead monorail 2). have each lever extend out over the monorail the other way and have a counter balance.
My first attempt to correct the bouncing would be a simple procedure. Place a secure point at the top dead center of the extruder (eyelet swivel). Tension string between the eyelet swivel and an overhead pulley. At the other end of the string a counterweight located in a remote area equal to the extruder mass.
It’s a beautiful machine you created. Best of luck to you.
I've never seen anything like this before! Cool! If I could guess it's either A: the bearings on the printing head aren't sufficient or there's slop in the joint B: there's something creating a perpendicular movement in the printer head joints which are causing the axel to buckle thus causing the bounce.
Thanks for your comment.
I will try to check your points!
I don't know how to solve the bouncing issue, but watching this dissolves all of my stress and worries.
@JKLee Try reducing the backdrive on the middle arm! If this arm could provide a little more control it might help stabilize the changing angular torque values applied by the outside arms! The shaking occurs when these two arms move, it seems! That geared arm you just showed off, should help!
try using shoulder bolts and properly toleranced bearings for the joints, make sure the bearings have a moderate press fit into the printed parts.
a normal threaded bolt won't have a precise enough OD to fit perfectly into a bearing like this.
A single axis of movement along the extrusion makes sense, but the head end should be ball joints to support full 3 axis of movement, on all three points. You are binding (bouncing) because the head end is restricted to a single axis, when really each point at the head end needs 2 axis of movement, the easiest solution would be to use ball joints on the head end which would allow 3axis per arm, where only two would be required, but with 3 you would not have to align/clock the two Axis per arm. The middle arm controlling y/z retraction can stay as a single axis movement. The outers need to be ball joints on the head.
This is really cool!
as some suggested stiffening up your structure is probably first thing I would do specially your adapters on the carriages, they should be pretty stiff. If alu is not an option, you can cover them with fiber and epoxy, like they do on boats :). Stepper motors have a lot of ripple effect, on a system that is not ready for it, it might multiply largely to cause all the bouncing one solution can be to use BLDC motors, but they can be pretty expensive or sort of a smoothstepper.
If there isn't a term to describe that design, I'll coin one "inherently non-rigid"
Here, you see the differnce between mathematicians and mechanical engineers.
Honestly even with 100% infill and high quality bearings in all joints, this will still not be a stiff design.. but low stiffness machines can be compensated with advanced motion control schemes. But thats a really deep topic..
I'd be interested in knowing how to solve this with motion control schemes.
@@nikolaivillitz6026 dude I'm on it give me a second I have a INCREDIBLE video
@@melody3741 Muahaha, I literally love niche technical videos on this sort of stuff, so send it on over :D
The printed parts can be steel wire reinforced as a potential option
Good point, but you shouldnt really care about these points in early Prototypes. The levers could be made out of aluminium or fiber reinforced stuff later on
Once you have implemented the other ideas, to reduce the effect of slack or binding in the joints whilst all the joint are under twisting force as far as dimensional and budget limitations allow, you could look at secondary considerations like
a. moving the COG of the head assembly directly under and as close to the intersection point of the three planes as you can to reduce the effect of the wobble;
b. detuning any resonance of the system by adding tuned dampeners like buildings use (especially when there is one significant resonant frequency) or using something similar to how a dead blow hammer spreads the force from movement eg build the arms with internal spherical voids and 1/4 fill with sand part way through the 3D printing and then complete the 3D printing to enclose the voids; [I thought of suggesting making the outer arms lighter than the inner and thus resonate at different frequencies, but you don't want to induce an action like a whip];
c. ramping up and down the stepping frequency and/or the phase of stepping so the effect of the each step is out of phase with the last and thus dampening the bouncing that is due to resonance;
d. and lastly, after you have done everything else, if some of the bouncing is due to joints slipping once an initial resistance is over come, deliberately inducing small jiggling/resonance in the stepping, even to the point of sometimes stepping in the wrong direction, to get or keep the joint slipping rather than stopping and binding may help.
I hope my suggestions help, from a software professional who enjoys thinking about hardware, but leaves it to others to spend the money actually making stuff ;)
Thanks for your well summarized comment!
I will try to implement head design - good to know COG under joint to reduce effect. Additionally, I will also consider stepping frequency due to Z axis and X/Y has different step rate.
Yes, there is many things are not optimized. But, your comments will help improvement. Thanks!
This is actually amazing. The simplicity of having one rail and the massive scaleability means this could be a very promising design for large-scale industrial FDM 3d printing. amazing work!
Video of optimized one has been uploaded.
ruclips.net/video/69kudlOXwMs/видео.html
Amazing, looks like a boss in a shootem'up video game 😅
Haha correct Tony
sound like a retro-boss too :D
I would guess this is a rigidity issue, if there is any flex in the arms or joints it will be used instead of the wanted motion, if the force required for the flexing exceeds the force required for the extruder assembly to move then the bounce will hapen.
To fix this you could try to counterbalance the arms so that the effective weight being moved is as low as possible and is always less than the stress that causes tortion.
Counterbalancing can be done by making each arm extend over it’s pivot point and adding weights to the back.
Completely removing the bound will be a balancing act between rigidity of the members and weight.
Lowering the acceleration should also help
If you take the belts off, a correctly balanced machine should stay in whichever position you place it in. Getting as close to that as possible should be the goal for this machine. Double bracing the joints might make them more rigid to counteract the flexing too.
Try replacing the members with 2020 extrusion for this extra rigidity, then double brave each joint
First of all, thank you for detail comment. I will try to make next arm based on your comment!
Immediate steps could include tightening the bolts holding the bearings and/or reducing speed/acceleration on the movements
@@AxfordIndustries thanks for comment. I tried to adjust tghting of bolts and speed/acceleration decrease. There is only changes its bouncing frequency.
@@jklee8866 also reduce jerk to zero if you haven't already.
I think the arms are too springy with plastic. Might need to try it out with carbon fiber rods or similar instead to increase stiffness.
Also as others have said you should probably have thrust bearings, at least for the middle leaning arm.
Before you can really smooth things out, you need thrust bearings with preload. The bearings you're using are suited for radial loads, but you're dealing with axial loads. Any slop in the bearings, which radial bearings have by design, is amplified in the print head. It may not fix it entirely, but looking at how smooth the linear rail bearings run, it doesn't seem like the controller, arm stiffness, timing belts or steppers are the issue. When in a fixed position, I suspect the print head has a lot of play in it. Thrust bearings with some preload can reduce this drastically. You need at least two pairs per joint though, and the axles for each joint should have no play. Precision shoulder bolts sized to the bearing would work, but they can add weight. Accurately sized pins with c-clips can reduce weight, but you need to size them properly to get the right preload.
Others have also suggested flextures, which is a great idea too. Using spring steel instead of pivots has a lot of advantages (like having your arm segments share the same plane). They can also give the linkages some preload. It can last too - the plastic arms will wear out before the spring steel if design right. Look up Dan Gelbart on youtube, and find his videos on flextures.
Thanks for your comment.
I also agree the possibility from controlling or stepper timing point.
So, I'm considering to modify that point.
Anyway, many people pointed out the two pair of bearing and preload.
I will prepare and share what the arm joint design.
You can do modifications of different tiers:
- tune the stepper profiles, implement active damping (low cost, just time)
- use stiffer materials for your design: 3D-printing dynamic load bearing parts is asking for trouble, i.e. replace the orange arms and the mounting bases (more cost, potentially outsourcing the manufacturing of the parts)
- scrap that design and opt for some kind of delta configuration (pragmatic for stiffness, but not as flashy). As another comment mentioned, there are inherent flaws with that design, most importantly that you can only get so much torsional stiffness out of a MONORAIL.
Very cool concept though, and I'm convinced you could get to something decent with the first two solutions
Video of optimized one has been uploaded.
ruclips.net/video/69kudlOXwMs/видео.html
We had similar issues with the original Morgan. The flexibility of the arms is too high, even if it is very small. To solve this we eventually replaced the arms with large diameter carbon fibre tube with the bearing mounts printed.
With this machine, even the smallest flex in the arms will cause a jump. Replacing your arms might need something even stiffer like aluminium alloy.
Fascinating movement though.
Correct me if I'm wrong, but it looks like these joints all work with limited range of motion, right? When I see limited range of motion + stiffness requirement, I think flexures. What if you could turn those bearing pivots into flexural joints? Easy enough to print, the design itself seems like the hard part. I've seen the double butterfly joint used before, though I'm not sure if it's the best for your situation.
Just an idea! Cool stuff, keep it up.
Do flexures offer repeatability? And don't they break after a certain amount of movements?
I always find those fascinating, but they gotta break after some use right?
@@MrRonny6 I don't know enough about fatigue failure in plastics, but yes, any material will fail due to fatigue eventually if the loads are sufficient. For this application, the loads are relatively low and the number of cycles to failure would be high enough to prove the concept.
Pre stressed ball bearings. Flat / axial bearings. Some bounce will be unavoidable given how long the rods are - all the slop, all the flex will be multiplied by the length... Still an amazing concept. Minimalistic, good for "general purpose" positioning... will be hard to make it precise though. Will require extreme rigidity and stiffness - or additional sensors and complex feedback loops.
Thanks for your comment.
I'm preparing explains of the bearing arrangement.
Actually, there is no clearance by preload of bearing arrangement.
The explains would be interesting to see.
Without clearance / restressing the ball bearings, it will have some wiggle in the joints. And friction will cause some "skipping", "jumps" - bumpy movement.
Another thing to consider is mass. Adding extra mass to the cargo will slow the thing down but - change the natural frequency of the structure too. The wobble can be made "slower" if the mass is increased. And if the cargo mass is reduced - the cargo will oscillate at a higher frequency.
The stiffer the rods and lighter the cargo - the higher the natural frequency of the structure. And using steppers - some oscillation will not be avoidable. Servo motors would be more smooth.
Everybody talks about making joints stiff and preventing binding and I'm wondering do we have enough resolution on the belt driven axes. We basically translate rotational motion (stepper) to linear motion (linear rail carriage) to rotational motion (arm) again. Any positional error of the stepper motor due to motor imperfections or microstepping is basically multiplied by arm length. From what I see the two outside arms (belt driven) should never move independently - one arm stationary when the other moves as it would bind the whole mechanism. As the arms must move precisely in sync so must the steppers. Any stepper positional error is translated into force stored into the belt (which acts like a spring) and released when positions get in "sync" again which is visible as wobbling, especially when doing translations on the axis perpendicular to the rail (only outside arms move).
From what I gather there are three ways to fix this:
1. Make arms driving mechanism stiffer (switch from belts to lead screws) so two steppers would basically force each other into position "in between the error" - this could lead to real binding and damage to the arms and/or head.
2. Make arms and/or head more compliant so the stepper positional error won't be stored as energy in the belts - this could probably lead to inaccuracies in the head positioning.
3. Put more positional accuracy into the system by using more precise stepper motors (0,9 degree), gear reduction (backlash!) or belt reduction (smaller pulley?) - this is probably the best compromise.
Anyway great job so far. I hope you have enough perseverance and helping hands to actually polish this project as it would lead to seriously cool and portable 3d printer.
Thanks for your comment. Your 3 point are fully agreed. I will keep in mind for next modeling. Thanks again for good summary!
The applications of potentially using this one day to simplify the printing of foundations to houses by potentially having something like this mounted on a long trailer, pulling up to a lot, and just “printing” with concrete is really cool to think about. This tech will be bigger than we can imagine one day.
Looks like most people have covered this already but the further the point you are trying to control is from the source of the movement the greater the amplification of any small variations or gaps. The arms are essentially fulcrums and any play in the material as well as the joints is going to cause what is seen here. You would need some vary rigid materials and some joins with very very tight tolerances to make this usable.
Looks really cool, and a its nice to see it proved from the renderings, but it seems that any tolerances are magnified a lot, belts, bearings and arm flexure ect, when you get it printing make sure you try a large circular shape and see how well it can manage that.
Fascinating gizmo, I could see this becoming a really long 3d printer when you get it tightened up. Not quite a belt printer, without the issues of a belt printer.
Bonus points for it sounding like a 1980s arcade. Makes me want to pop a quarter in the Asteroids machine.
This is by far one of the most fascinating thing I've seen moving.
Looks like @Kelvin Nishikawa has the best idea for fixing this with the thrust bearings. I also thing maybe if the arms were something more rigid, like cut aluminum, though that would be a challenge to acquire.
Remember that a robotic design with some flexible "springiness" inherent in its construction can be safer in close proximity to humans; "softer"(a virtue) in that it is less likely to have high-G-rate(damaging) crashes. Depending upon how it is used and what applications it is used for, this might be a virtue not a vice ==>Interesting mechanism, may have many good applications, food service first came to my mind.
Joint free play control, and how it affects/effects overall unusual motions would be where I would usually start(but this mechanism is new to me)
Kudos on your work, (subscribed✔️). As I read more about the kinematics and your intended function, I can be more helpful for solving your problem,,, just to remind you that correctly used(adjusted) the fact that there is"bounce" may be a good thing (tuning it down to the correct amount for your application, is the question)
Again, interesting stuff, well done👍
My recommendation would be to put all arm mounting points in double sheer. I don’t think the rigidity of the plastic is causing your oscillating bouncing. I think when the motion changes direction it moves the load on the pivots from one direction to the other and the pivots are flexing slightly because you’ve turned all pivots into a moment arm. Having them all in double sheer I think would improve the machine.
Thanks for your comment.
Especially Z axis arm, I agree with your point!
I will consider double sheer design in next design.
It looks like static friction is the problem. As the arms move they are building forces but not enough to move, and when enough force is applied to overcome static friction it snaps, like an earthquake. I would recommend disconnecting the motors, and movie it by hand to see if this is the case - follow a similar path and see if that is the case. Normal bearings can cause this effect too because of slim sleeves able to have too much pressure applied and end up deforming them slightly for cheap bearings or bearings using different material types.
You'll be able to feel where the problem is - our hands are incredibly sensitive, so when a snap happens, just try to follow that - or have someone else movie it and you touch the arms, etc.. and feel where those jumps are.
Edit: And bearings do have gaps - so it may move a little and then impact requiring move force to move or rest on the wall already.
The attachment of the middle arm is causing some of that by binding against the motion of the other two arms; the other person who mentions thrust bearings is right, but you also need some sort of counterweight, the weight of the head is more than that of the arms and they're not rigid. If you don't want to redevelop the center arm so it's not binding as much against the motion of the other two, I'd look at some sort of counterbalance
And speaking of this in terms of robotic arms and stuff, usually there's springs, counterweights, etc, to reduce that mass shifting or flex in the armature.
What I'm intrigued with your implementation there is how you're driving the arms with the single rail. Interesting choice for a limit switch, by the by. It's a bit of a deviation from the typical microswitches or optical solutions.
Amazing that it works at all. It seems like this would need extremely rigid joints and arms to work well.
As has been noted, the lower arms are mostly loaded in torsion. Box the arms (ie. add multiple top layers) rather than the single face or try manufactured tubes with printed sockets.
This is an awesome and fun proof of concept!
To be honest the level of bouncing you are getting is pretty good considering how minimally constrained this is - the monorail setup is placing the maximum possible demand on every single joint, bearing and plastic part. I believe the main source of the bouncing is normal 'ringing' from the belts, but able to resonate through the many joints fighting one another.
Tighten the belts a bit more but not ridiculously tight. Thrust bearings could help. Lighten the arms further - carbon fibre square tube or thinner CF strand reinforced plastic would be good. Put dampers on the stepper motors, and this might sound a little mad but a little vial of liquid on each of the moving carriages might help absorb some vibrations.
Not sure what's reasonable to expect quality wise from a 3D print from this. but I hope you can get it to functional!
your toolhead is bouncing due to binding, either in the bearings themselves due to off-axis loading, or the segments of the links rubbing against each other. You could try using a small amount of a PTFE bearing silicone lubricant between the segments where they are rubbing and see if it decrease the binding. Another thing to try is doubling the amount of bearings between each segment, which should help reduce the side loading on the bearings.
The other consideration is the segments in the arms themselves twisting, since you are relying on the arms being rigid members.
Ultimately, while I like the idea of this kinematic system, I feel that it likely has a lot more bugs to work out than the bouncing you are getting at low feedrates.
Good luck!
a couple of ideas that come to my mind :
- i believe you move the axis in curve (non-linear steps) maybe you can refine your curve more with even finer steps?
- use higher bit microcontroller, eg. used 32 bits instead of 8 or 16 bits
- make use of ball joints instead of "elbow" joint
anyway @JK Lee great jobs! hope to see more! *following
Hello,
Have you got a useful answer?
Here is a list of possible causes of this bouncing:
1 - Rigidity of the material of the arms.
2 - The rigidity of the support of rail on the table.
3 - The play in all joints of the arms, and how the ball bearings do not play in their holes.
4 - The precision of the guide on the rail.
5 - The elasticity of the belts.
6 - The precision of the screw and the play of the nut.
7 - The way the screw is maintained along its axis and the coupler. You cannot rely of the motor rigidity nor on a soft coupler even though you may need one.
I do not see o the videos how the screw is mounted at each end.
This kind of nuts is a sacrificial part of 3D printers that force only in one direction on the printers. Some printers incorporate a system composed of a spring and a second nut that constrain the main not to always be in contact with screw. But this works is the nut is acting always in the same direction.
The first thing you could do is to make static measures.
Block the motors in one position where the head is 100 mm away from the rail for example and maintain the motors (do not disable the drivers).
With a comparator you will put a bit everywhere to measure the displacement while applying manually a force, and try to evaluate
- the play on your system (slightly pulling up and down by hand, feeling when the system is constrained in one direction and the other)
- the stability of the thing, by forcing a bit more.
Regards
Accumulated play thats perpendicular to the axis of rotation of each elbow bearing, twist. You need bearings with some preload rather simple pivots, small angular contact or taper rollers ideally to handle both load directions. The arms need to be very stiff in the bending moment, not simple beams, plastic is not ideal.
The bounce is due to the arms negotiating stray torque forces.
I assume the arms always maintain a air gap and are linked only by the inner race of the bearings minimizing the chance of stiction (stiction would cause a similar issue but unlikely at this stage of the project).
When you have a device with two arms they tend to self balance the minute differences in stray torque with one taking the lead in moving and the other following, but with 3 or more arms they will bounce due to force being loaded on one arm until it bounces free and then force loads on another.
If you had a perfect mechanism with Idea servo's and magic math you wouldn't get stray torque but in real application its impracticable to implement a level of accuracy to eliminate stray torque so the best bet is to accommodate it.
To reduce it effects you could introduce a ball joint on the rail end of two of the arms the arms or the outer two cradle ends of the arms.
Add a rubber band between the end point and the arm connection on the rail. This will keep a more constant pressure against the bearings and should smooth it out. Like shock absorbers do.
There's a reason tripteron is not a commonly used design, and that's also why SCARAs look surprisingly beefy. Joint on joint on joint can be a source of flexing. This machine needs more stiffness. My suggestions:
First, use much larger bearings (3x dia. or more) and preload them. It will increase resistance, but it's a trade-off. You don't need thrust bearings. Large, thin deep groove type is probably fine.
I'm sure the arms are twisting. Since it's mostly the perimeter that takes this kind of load, you can make them larger but hollow to keep weight low.
Thermoplastic alone is not the best material in this case. Try to reinforce them with a fiber composite shell.
Tripteron kinematics is mesmerizing and your monorail version is a beautiful take on it. The practical implementation will be a challenge though.
What I see is that when your two forward arms are extended too far in the x direction along the track, you are getting a small amount of twist though those same arms. When the torsional pressure is released from the extension of the arms is when the bounce seems to occur. Y-axis rotation at the base of the two forward arms may help solve that.
I think the weakness of this system is that it relies on the arms and bearings taking forces that are out of their movement plane. This causes twisting torque on the joints that increase friction, binding (alternating between rest and dynamic friction) and potentially slack. You would need more robust multi row bearings and joints that can handle this well.
After watching this mesmerizing video for a minute or so... Looks like compliance (material is bending) in the 3D printed arms. I think I noticed this mentioned in some other comments. Look up Delrin, its a fairly stiff and very machinable plastic that could work well for you.
it's the pulses in your signals, you hear as different music notes, being amplified by the long arms
if the signals to the steppers motors changes frequencies more smoothly, then the bouncing would go away
it's a combo of the harmonies between the different stepper motor speed changes
MY humble idea on your bouncing prob.
: Arms having intermittent frictions issues with each other at the rotating joints because there is sideways forces pushing them at each other. But wachers to separate them.
To fix up and down oscillations, The center arm needs a second arm operating semi-symmetrically on the same rail. Kind of like a really complicated scissor-jack. Basically, that center arm, needs a triangle with 2 points on the same rail.
I think your right or on to something at least.
I feel like if you changed the joints at the end of each arm to accommodate more angles of movement, such as a limited ball joint.
I think it has to do with the bearings in the middle arm being twisted into angles it doesn't like, giving that weird effect. Might need higher quality/bigger bearings, maybe even a different style
After more thought adding a bearing to each joint, it would change the center of rotation to the walls of the bearing giving it a more 'normal' load instead of twisting them.
The problem resides in the stiffness of the joints. They need to be extremely rigid and precise with no backlash. They also need to be fluent and friction free though. Another problem might be the rigidity of the arms, if they are too flexible the bouncing effect is amplified.
Possibly longer linear bearings and associated plastic mounts. I suspect the twisting forces here are the largest contributor to loss in rigidity. Cool machine.
This would require a pretty rigid setup with thrust bearings and perhaps metal arms
There are many solutions which you probably already plan to implement on the next iteration, but just in case, I will describe them: Oversized the cross section of the arms and print them with a stiffer Carbon Fiber fill filament. A professional version of this would ideally have cast Aluminum arms for rigidity. There is just too much cantilevering going on and every movement has a vertical component to the forces exerted. This is the nature of the thing with those 45° mounted bearings. Every effort must be made to make this thing as lightweight as possible. Lowering inertial mass will reduce the bouncing. With this much undesirable slop/motion as shown in the video every technique possible should be implemented. For instance: Reduce the weight especially out at the ends of the arms. What's the lightest weight Hotend you can find? Also, you could upgrade the bearings to cone thrust bearings but this will be expensive and difficult to source the parts. I believe you can get acceptable performance from the 608 Skateboard bearings but you will need to space them with a larger gap for more rigidity. Each end of each arm should have a pair of bearings mounted maybe 30mm apart. For a total of 4 bearings per joint. Your design will need a way to adjust the axial preload force on the bearings. This will eliminate slop/backlash in the joints. Good luck and keep up the fantastic work. I love this project!
Thanks for your good summarized comment!
@@jklee8866 Speaking of lightness and rigidity have you considered making an arm from two printed parts, at each end for the bearings and one middle part using one of those standard premade carbon fibre rods
Pretty sure it needs to weigh more not less and have increases joint stiffness by using aluminum that is thicker, so it cant twist
The monorail design puts a lot of strain on all joints and bearings. I believe that was obvious, when you designed the mechanism. So you need to address this as best as possible. A first and probably easy step would be to double the orange arms attached to the holders on the rails. This might even be done without any design change. At least for the left and right arms. Just lake longer screws, and add an opposing arm with bearings. That should increase the rigidity drastically. Looks like it might fit. Maybe even also double the secondary arms connected to the head. That would mean 4 bearing on one axis and way more stability.
Thanks for your comments. Especially, the arm of Z axis is week arm to bear the load by head. I will consider two arm or wider arm as your comment. Thanks!
Just tossing this out there in case it's helpful, I do think the others are right about binding, but if that *doesn't* solve the problem, I would look into the controller. This thing has to do some pretty weird trig/IK to put the effector where you need it, and I know in delta printers you can get issues if your controller is a bit slow. It's not likely the cause, but it is something to look into if the problem sticks around.
I will also consider controller side as your comments. Thanks!
The problem is in the geometrical design of the machine, there are long plastic pieces that hold the extruder that have some flexibility, also the bearings allow a little movement, and the geometry of the machine makes all the play and flexibility to multiply, a more conservative design with screw rods in the xyz axis will have less vibration.
@JK Lee: this could be because the balls and the running surfaces in the ball bearings are not round enough. The ball bearings are available in different precision classes. The more precise the more expensive - of course.
It's probably the actuators resistance to each other. As an example when the leftmost one is going right the middle one has to fight that force which is being applied to it by the fact that they are connected all together. That produces bounceback movements and overshoots that are not accounted for by the control system. I have no clue on how It works. Maybe if you are using a PID controller try to raise the I term.
But then, i might be completely off.
To test this maybe try to hard lock and fix mount two of the actuators at a time and just move one and see for each of them if you still have bouncing problems, once you get rid of them by isolating the single movement try to account for the forces applied to and by the other actuators
Try using aluminum rectangular tubes and axial bearings combined with radial ones for the linkages. You should be able to reduce the vibrations for a bit.
Hello, my guess to improve this:
Assembly is not steady enought due to (1) too much axis and (2) weak motor power to fight gravity.
-maybe try with elbow type joints like crane (but not wired, with hardware materials) and grease on elbows
Or add motors on joints to help, but this will increase weight...
Anyway you have althought already a beautiful machine!!
seems hothead looks to be heavy for arms size.. tip end placement increases leveraged weight too.. so, lighter massed hot unit or stronger wider arms might distribute end weight easier to carry hold steadier? wider contact (perhaps polished steel washers few cm's larger) by bearing rotors joints would align movement by surface contact glide opposed to lifting strength alone.. amazing compact design concept for dimensional movement.. nice work! keep at it..
hmm looks like friction and non-smooth joint movement!
amazing work! very inspiring.
I would like to point out that the arms aren't perfectly designed nor is the material strong enough to compensate so it might be the elasticity of the material just shaking the extruder.
this issue is compounded by the rough movement of the steppers, I recommend trynamic to at least have a smooth movement, well lubricated rail, joints and nut and screw. If the issue persists try changing the arms with some carbon fibre or steel rods, perhaps you could use pipes instead of rods or multiple rods in parallel to form a single arm.
I think that this solution where the head is placed at the end of something equivalent to a rod will always have some "flex" in it and vibrations will magnify. It also will have a natural oscillation frequency. As others mention to stiff things up will help.
While I don’t operate 3D printers, it appears that the bouncing is coming from a bending moment between the mid and head bearings. I would recommend that you thicken the mating area by the head and implement a long set of needle bearings. Thrust bearings also seems like a good place to try first though.
The circle motion sounds like music. I feel like the motor holds a certain speed for a short while and a moment later it changes it speed with a little bit instantly. It doesn't sound like a gradual change of speed. Maybe these constant shocks through the arms result in a bit of bouncing too.
Of course this is just a hypothesis and it could be wrong, but if this is not the issue (or one of the issues) then at least you can be more sure of the right answer :)
Suggest you look at the Tripteron design done by Aspu there. The monorail idea is intriguing, but the basic simple joints you're using there lack sufficient stiffness along with the printed arms (Aspu started out with 3D printed arms and then moved to printed hinges, etc. and 2020 rails, etc. ) which is where at least part of where the bouncing is coming from.
1: Conical ball bearings so you adjust away ALL play
2: Extremely stiff arm-parts.
Cool idea!
But everything needs to super-mega-stiff and totally over dimensioned for the load.
I'm not an export ..
But you may check "basics" points :
- dose all arms move freely ?
- did you see friction on each parts ?
-> in fact the motion look jerky only on one direction
=> did you try linear motion 1st ? (only X / Y / Z and check if you are able to have linear motion - check "speed" too to see if you have speed variation -> set only one axes and record the motion then check frame by frame ... the other method is to draw the line and move the paper in the other way -> you must have line to but if you have a curve, it's means the speed is not constant according the angle. Camera record could be easier to test )
As I can see , the orientation of each arm are not the same, so it can be a speed issue (you have 3 motors but you have 2 arm axes so one arm must move faster / slower as the others)
=> could you try to add a pen on the head ? I'm pretty sure you may not really circular
as I can see on the lase sequence (2:20 side view) the Z change but (for me) it shouldn't
-> this is a very promising prototype !!
I suggest to change two of the end effector bearings with ball joints, as the current design requires the rotation axes per link to be accurately parallel.
The first thing I would do is add z-bend to the 3 "forearms" so that the forearm facing sides of the bearings on printing head were in the same planes as bearing faces on the rail brackets (right now these planes are offset by the thickness of the forearms). Next thing is to locate the points of intersection between those planes and corresponding axles of the rail bearings. All three points should be on a same straight line parallel to the rail (right now the middle point seems to be too far back, and I am not sure if side points are aligned either). These changes should eliminate a lot of small but randomly directed forces on the bearings. If it is still not satisfactory, I'd replace ball bearings with roller bearings.
You could try a finite-jerk motion planner.
Or, possibly, just use links that are stiffer.
When theory and practice differ, it is frequently because you have forgotten that everything physical is a spring.
Thanks for your comment. I fully agree with you. Steel show bending in some heavy applications. Motion wise, I think the force equilibrium problem.
Amazing. I guess the connection to the rail might be weak and the material you are using also might add some movement to it. If is not too much trouble you could try with aluminum arms and stronger joint. Amazing mechanism though 👏👏👏
I'd recommend changing the arms to interleaving "fork and blade" connections at the pin joints so they can't put as much off-axis torsion on the bearings.
Perhaps think about some type of suspension between the arms such as some springs or wire to keep the frequencies of any movement/vibration even throughout. Shouldn't eliminate shake but could smooth movement.
Besides what others have mentioned regarding the stiffnes of the arm I think there could also be an issue with the trajectory.
Basically If one motor has to travel a short distance to reach the desired position and one has to travel further, the second one is gonna be there later. This can result in undesired pathing of the end effector. even though the position in the end is correct.
You can test the mechanics by just driving one axis at higher speed. Break with a high deceleration and see how it behaves. If it bounces after that you have a problem with the stiffnes of your arms and joints. If the problem prevails even at low endeffector speeds it is most likely a problem in the trajectory generation
Very nice mechanism for a 3D printter!
I think you must attack the vibration souce, try to reduce at minimum the vibrations. Apart from that you can always improve the joints quality or matemathically model your system and obtain from the model the value of the optimal dampers, test some of them and so damping the system.
But from my perspective, just watching the video cant tell you more than it looks like your are using stepper motors and the jump between steps is to big and agresive and so produce the vibration that is amplified by the lenght of the arms.
You can try to use shorter arms to see if the vibrattion is reduce and so valifate this idea.
If this is the case... the are several options!
Seems to be due to friction in the bearings at small angles from arms 1 and 3. The arms are pushing against one another and its likely causing positioning inaccuracies due to friction lag. Perhaps consider adding an adjustable tension rod between the two arms to make them more connected and stable!
You have a lot of lateral force on the arms. If you watch the center arm closely, you can see it's slightly twisted due to the force, and the joint is being pried open. The materials aren't stiff enough and the joints aren't rigid enough, so it's bouncing
You could make the rail effectively wider by installing a second rail and set of sliders a few centimeters from the first, and attach the arms to both rails with a linear bearing on each rail. That'd give you a longer moment arm to resist the torque moving the head up and down.
to me it looked like the arm attached to the monorail wants to tilt at the pivot point but it binds try to give the arms connected to the monorail some room to tilt right to left and try to clamp the ball joint on both sides i'm no expert soo take what i write with a grain of salt but hope to see more update videos it's fascinating good luck man
Looks like the normal play in the bearings are being amplified by the long arms. Could try tapered roller bearings, much tighter tolerances due to their design.
A cheap band aid fix could be just add springs at each joint to add tension and smooth out some of the bounce.
I'd move away from plastic parts and not use bolts on the joints to start with, the belt is likely not helping either.
Very cool mechanism.
I think the geometry requires the arms to be very rigid and for the bearings to have no free play whatsoever. The bouncing seems to happen when large movements of the head are being produced for very small movements of the drivers.
I think just making everything more substantial would solve the issue.
The geometry and math to make that move seems fascinating!
I'm also feel like you when I first watch Nicholas Seward's 'RepPap Monorail' concept.
ruclips.net/video/oBBhVcemVHc/видео.html
This is what I try to make it real based on his concept.
@@jklee8866 Neat!
Just wow ! Despite all the possible limitations, it's just wow !