Sticks, Slips and Rusts: Designing 3D Printed Mechanisms That Move on Purpose

A printed mechanism is a strange thing to build. Every part is soft, every joint has play in it, the pin is not perfectly round, and the hole it turns in is not perfectly round either. Yet a printed hinge, a printed plier, a printed latch and a printed parallel arm all work, and some of them work for years. The interesting question is not why printed mechanisms are hard. It is why the ones that work do work, and why the ones that fail almost always fail in one of four ways that have nothing to do with strength.

This article is about motion, not structure. The previous pieces in this series dealt with parts that had to hold still: load paths that do not creep apart, gears that survive a real load, clamps and hooks that hold on to something. A mechanism is the harder case in one specific way: every part that must move is also a part that must be located, and those two jobs pull in opposite directions.

Everything below assumes the parts are already connected. If a joint still has to be designed, the companion pieces are pins, pivots, shafts and bushings for the connection itself and compliant mechanisms for the case where the best joint is no joint at all.

Four ways a printed mechanism fails, and only one is strength

When a printed mechanism stops working, the owner usually assumes it broke. It almost never broke. It moved in a way the design did not intend. Sorting the four failure families is the first useful thing you can do, because each one has a completely different fix and three of the four cannot be fixed by adding material.

Failure family	What the user reports	What is actually happening	Where the fix lives
Seizure	It worked, then it went stiff and then it stopped	Two surfaces pressed together until the plastic welded to itself	Clearance, contact pressure, material pairing
Seizing by debris	It grinds, then it jams	Wear particles from a printed surface have nowhere to go	Geometry of the gap, not the size of the gap
Overconstraint	It binds at one angle and is free everywhere else	Two pivots fighting over where the link is allowed to be	Degrees of freedom, joint order, compliance
Load reversal or impact	A tab or an ear snaps off	The pin is loaded in the one direction a printed layer cannot take	Orientation and boss design

Notice that three of the four produce the identical sentence from the user, which is roughly "it got stiff". That is why diagnosis has to be done by disassembly and inspection rather than by reasoning from the outside. A mechanism that bound at one angle has polished witness marks on one side of a bore. A mechanism that seized has a dull, smeared, slightly melted-looking track all the way around. A mechanism jammed by its own debris has grey dust in the gap between the parts and a clean bore behind it.

The clearance nobody actually measures

The single most wrong number in printed mechanism design is the radial clearance in a pivot. Most people pick 0.2 mm because they read it somewhere, or 0.4 mm because 0.2 mm seized once. Both numbers can be right and both can be wrong, because the number that matters is not the clearance you drew. It is the clearance that survives printing.

Three things eat the clearance you drew, and none of them is visible in CAD.

• Hole shrink. A hole printed in-plane comes out smaller than drawn, typically by 0.1 to 0.25 mm on diameter, depending on material and wall thickness. The pin does not grow, so the whole error lands on one side of the fit.
• Layer-plane scallops. A round bore printed with its axis parallel to the plate is not round. It is a stack of chords. The effective reduction is roughly one layer height, which at 0.2 mm layers is a tenth of a millimetre missing from the diameter before anything else happens.
• Seam and z-offset swell. The seam line and the first layer after a travel move sit slightly proud. On a bore, that proud material is a pair of ridges that touch the pin at two points and nowhere else.

The practical consequence is that a nominal 0.2 mm radial clearance is closer to 0.05 mm on the tightest side once the part comes off the plate. That is a press fit, not a bearing. If the mechanism is meant to turn by hand without a bearing, start at 0.35 mm radial on a 6 mm pin and test, and plan to print the pivot in the orientation that puts the bore's axis vertical so the bore is a true circle rather than a stack of chords. The pin, in that case, gets printed lying down for the opposite reason.

There is a second reason to keep the bore axis vertical: a vertical bore can be printed without support and without a bridge over the top of the hole, and it will be round to within a layer height. It also fails in a predictable way if it does fail, because the weakest direction of a printed tube is straight up the layers, not across them. For a pivot, that is the direction the load usually does not push.

Contact pressure, not clearance, decides whether it seizes

Clearance decides how much the mechanism wobbles. Contact pressure decides whether it survives. The same 0.3 mm clearance will run dry forever on a lightly loaded idler and weld itself shut in an afternoon on a pin carrying the full reaction force of a loaded arm. The load does not distribute over the projected area. On a printed bore with a slightly oval hole and a slightly out-of-round pin, it distributes over a contact patch that is a fraction of the nominal area.

For a pin of diameter d and length L carrying a radial load F, the nominal bearing pressure is F divided by d times L. That number is a useful first cut, but on printed parts you should treat anything above roughly 2 MPa as a design flag, because the real contact patch is much smaller than the nominal one. A 6 mm pin through a 10 mm long boss at 2 MPa is 120 N. It does not take a large lever to produce 120 N at a pivot. This is the arithmetic that explains why so many printed arms seize at the root pivot and run fine at the wrist.

Situation	Rough bearing pressure	What to expect	Reasonable response
Idler, guide roller, cosmetic swing	Under 0.5 MPa	Runs indefinitely dry in most materials	Nothing. Leave the clearance alone
Hand-operated lever, latch, plier	0.5 to 2 MPa	Runs, slowly polishes, may develop a flat	Longer boss, or a steel pin instead of printed
Motor-driven linkage, cam follower	2 to 5 MPa	Seizes or wears oval within hours	Real bearing, metal pin, or reduce the lever arm
Any pivot carrying an impact load	Above 5 MPa instantaneous	Turns into a hammer on the second use	Compliance upstream, or do not print this joint

There is a cheap way to buy a factor of three here without changing the material. Lengthen the boss. Bearing pressure is inversely proportional to length, and a boss that is 20 mm long instead of 8 mm turns a marginal joint into a comfortable one at almost no cost in print time. The failure mode of the longer boss is also gentler: it wears slowly rather than seizing suddenly, which gives you warning.

Two pivots are a constraint; four pivots are a fight

The most instructive way to understand why printed mechanisms bind is to stop thinking about parts and start counting what each joint takes away. A pinned joint removes two translations and leaves one rotation. Two pinned joints on the same rigid link, with parallel axes, remove two rotations from the link's ability to move freely and leave it able to swing only in one plane, which is exactly what you wanted, provided the two pins are exactly parallel and exactly the right distance apart.

They are not. The printed part has a pin spacing of 80 mm plus or minus 0.2 mm. The base plate has the same nominal spacing with its own error. The two errors add. Now the link cannot be in both places at once, and the only way it can assemble is by bending or by stretching the gap, which means the whole mechanism carries a permanent internal load even with nothing hanging from it. That internal load is why the mechanism feels tight at one end of its travel and loose at the other.

This is overconstraint, and it is the single most common reason a printed mechanism "works but feels wrong". The fix is never to open the clearance everywhere. Opening the clearance hides the symptom and introduces slop at the joint that did not need it. The fix is to organise the joints so that only one of them fully defines the position and the others are allowed to comply.

Arrangement	How many joints fully locate	Behaviour under print error	When to use it
Two parallel pins, both nominal	Two, so they argue	Binds at one extreme, free at the other	Only for very short spans
Two pins, one with a slotted hole	One, the other floats along the slot	Runs smoothly, small play in one direction	Default choice for parallel links
Two pins, one elastomer bushing	One, the other allows shear	Runs smoothly and centres itself	Where you also want vibration damping
Four pins, two parallel arms	Four, so it is a fight	Almost always binds unless one arm is compliant	Only with a deliberate flexure in one arm

The slotted hole is the cheapest fix in printed mechanism design and it is badly underused. Drill nothing. Model one of the two holes as an oval 0.8 mm longer than the pin on the axis of the link, print it, and the mechanism will assemble first time and run without binding. The play it introduces is one-directional, along an axis where the link's motion already has no consequences. That is a very good trade.

Orientation decides the load case, so decide it first

Printed parts are not equally strong in all directions, and the ratio is not 10 percent. Across the layer interface, the same material is typically two to three times weaker in tension than along the extrusion line, and far less tough. On a pivot boss, that matters twice: once for the hoop stress around the bore, which is carried by the perimeter loops and is fine, and once for the ear or tab that holds the boss, which is where the failure actually happens.

A boss torn out of a plate is a delamination, not a tensile failure. The pin pushes on the inside of the bore, the bore pushes on the material around it, and the material around it tries to separate from the layers below. If the boss is printed as a cylinder standing on a plate, the layers are horizontal and the load is horizontal, so the load is being carried across layer interfaces over the whole height of the boss. That is the worst case, and it is the one almost every printed CAD file produces by default.

The alternatives are few and all specific. Split the joint so the pin sits in a part printed on its side, with the load running along the layers rather than across them. Add a fillet at the base of the boss, if the load is diagonal, which it usually is. Sleeve the boss in a printed collar whose own layers run at right angles to the first part, so the two sets cannot fail on the same plane. Or accept the orientation and reduce the load by adding a second boss on the far side of the plate, so that each boss carries half the reaction and the moments partly cancel.

For a mechanism where the pins are the wear parts and the arms are the structure, this leads to a useful split: print the arms flat, print the pin bosses as separate collars turned 90 degrees, and glue or snap them in. The collar is the part that takes the beating, so it is also the part you want to be able to replace. The joint geometry for glued assemblies matters more than the adhesive here, and if the collar is held by screws rather than glue, the rules for screw bosses that do not crack apply directly.

Four linkages and where each one belongs

Most small printed mechanisms are one of four types, and the choice between them is decided by the motion requirement, not by preference. Being able to name the type you need saves a lot of printing.

Linkage	What it does well	What it cannot do	Printed-parts caveat
Four-bar (crank-rocker)	Converts continuous rotation into a reciprocating swing with a controlled path	Produce a straight line without careful link ratios	Needs a slot or compliance at one joint, or it binds
Slider-crank	Turns rotation into a straight push over a long stroke	Take side load at the slider	The slider is the wearing part; make it replaceable
Parallel arm (pantograph)	Keeps a tool or a platform level while it travels	Tolerate unequal arm lengths	Four pins, so treat as overconstrained by default
Toggle (over-centre)	Holds a position with no holding force once it has passed the line	Release itself without a deliberate input	Great for clamps, terrible for anything that must not lock

The four-bar linkage is worth a paragraph on its own because it is the one most people attempt first and the one where printed tolerances hurt most. A four-bar is defined by four link lengths and the order they connect in, and only some combinations will actually rotate continuously rather than jam. The classic condition, that the shortest link plus the longest link must be shorter than the other two added together, is easy to satisfy and easy to forget. Failing it does not produce a slightly wrong motion. It produces a mechanism that goes partway and locks, which the builder then diagnoses as a clearance problem and opens up, which makes it lock somewhere else.

If you are building a four-bar with printed links, build in a long slot at the coupler pin and a slot in the crank. Two symmetric slots cost nothing in print time and they let the mechanism find its own path instead of demanding that your CAD be perfect. A snap or press fit at the coupler is also handy, because it makes the link lengths adjustable after the fact, and adjusting a link length by 1 mm is often the entire fix.

Dead points, toggle positions, and why a printed latch pops open

A toggle is a mechanism that reaches a position where the input force no longer has any lever arm. At that moment, the mechanism can be held closed by a very small force, and it can also be pushed out of position by a very small disturbance in the right direction. Anyone who has printed a latch that holds beautifully and then releases itself in a bag has met the second half of that sentence.

Two geometric details decide which of the two behaviours you get. The first is how far past the centre line the mechanism travels. A toggle that just reaches centre is on the edge, and will release with any vibration at all. A toggle that goes 2 to 5 degrees past centre is stable. The second is the stiffness of the frame the toggle pushes against. If the frame flexes, the toggle only appears to go past centre; the energy is stored in the frame, not in the geometry, and the latch releases when the frame relaxes.

That is why a printed toggle clamp should be built into a frame it cannot bend, and why the frame is usually the part worth printing in a stiffer material than the lever. A lever in PETG and a frame in a fibre-filled grade is a much better pairing than both in the same soft material. If the frame is ASA or a glass-filled grade, the toggle position will hold for a long time, because the frame is not quietly relaxing under the preload.

Materials: choose by what moves, not by what is strong

The material table for mechanisms is not the material table for structures, and using the structural one is a reliable way to build a mechanism that binds. For a moving part the ranking criteria are surface hardness, friction against itself, creep under a small constant load, and how the material behaves when it wears. Tensile strength barely appears.

Material	Pivots and sliding	Wear behaviour	Creep under preload	Best use in a mechanism
PLA, PLA+	Poor. Wears fast and gets sticky	Produces fine dust that then jams the gap	Poor, fails slowly under any preload	Jigs, mock-ups, one-off throws
PETG	Good. Low friction against itself	Smears rather than powders when overheated	Moderate	Levers, handles, low-load pivots
ABS	Fair. Harder surface than PETG	Chalky dust, abrasive	Moderate	Frames near heat, snap-action parts
ASA	Fair to good	Similar to ABS, slightly tougher	Moderate	Outdoor mechanisms, latch frames
PETG-CF	Good. Stiff where it matters	Abrasive to itself; avoid CF-on-CF pairs	Good	Long links, arms that must stay straight
PA6-CF, PA612-CF	Good, with a real bearing surface	Tough, resists chipping	Good when dry	Loaded pins, high-cycle joints
TPU 95A, 85A	Not a bearing material	Self-lubricating and forgiving	Poor under constant load	Bushings, bumpers, anti-slip pads

The temptation is to use PETG-CF or PA6-CF for every moving part, because they are the "engineering" materials. For structure that is right. For a pivot pair it is a mistake: carbon fibre is abrasive, and two carbon-filled surfaces running on each other behave like sandpaper on themselves. The correct arrangement is a stiff composite for the structure and a softer, self-lubricating material for the contact surface. A TPU 95A bushing inside a PETG-CF arm is a better bearing than two PETG-CF parts on each other, and it is quieter.

Two more pairings are worth knowing. PLA on PLA is the worst common combination for a pivot that has to move: it is the pairing most likely to cold-weld under pressure, and the one that generates the most jam-producing dust. And any fibre-filled material on itself, dry, will wear the mating surface rather than wearing itself, which means the cheap replaceable part must be the hard one, not the soft one. That inverts the usual "sacrificial part" logic and is worth writing down before the design is frozen.

Lubrication, and the part where most advice is wrong

Printed mechanisms do not usually need lubrication. They usually need better geometry. Grease on a badly constrained mechanism turns into a paste with its own wear dust and makes the binding worse, and petroleum-based grease on some plastics causes stress cracking at the very joint it was meant to protect. If a mechanism needs grease to move, it is telling you the contact pressure is too high or the clearance is gone.

Where a lubricant genuinely helps is on a high-cycle, low-load joint: a printed hinge on a lid, a small idler, a snap latch used a thousand times. There, a dry film is the right choice, not oil. PTFE dry film, a silicone-free dry lubricant, or a small smear of silicone grease will all extend the life of a PETG or ABS pivot by a large factor, and none of them attracts dust the way lithium grease does. Apply it once, run the joint for a few minutes, wipe off the excess, and leave it alone. Re-application is a sign the joint is wearing rather than sliding.

There is one printed-mechanism oiling rule that is worth ignoring: never lubricate a surface that is supposed to grip. A pivot that must transmit torque, a press-fit collar, a friction detent, all lose function the moment a film gets in. The same instinct that makes you grease the pin makes you grease the hub, and the second one is a mistake.

Print-in-place: what actually determines whether it frees

A print-in-place mechanism is a mechanism whose clearances are decided by the printer rather than the designer. That sounds fragile and it is, but the ones that work reliably all share the same three features.

• At least three layer heights of gap. A 0.2 mm layer needs 0.5 to 0.6 mm of horizontal gap before the two parts reliably separate, because both surfaces sag into the gap and the elephant's foot at the bottom is wider than the nominal wall. Anything under 0.4 mm is a coin toss.
• No horizontal overhang inside the gap. A gap that is bridged over the top of the pin will close there. The gap has to be open to the plate for its whole height, or the top of the pin will be fused.
• First layer flattened. The elephant's foot is the usual cause of a print-in-place part that is free at the top and welded at the bottom. A 0.1 mm first-layer height difference in the slicer is worth more than any post-processing.

Print-in-place fails most often not because the gap was too small but because a circular bore printed with its axis vertical gets a slightly smaller effective diameter at the bottom than at the top, so the joint is welded at one end and loose at the other. Biasing the gap toward the plate, by making the lower half of the bore 0.1 mm wider than the upper half, is a small modelling trick that removes the most common failure.

A PEI build plate matters more here than for ordinary parts, because a slightly uneven first layer shows up directly as unequal gap closure. A plate with a worn or inconsistent coating will produce print-in-place parts that free on one side of the bed and not the other, and no amount of slicer tuning fixes that.

What to test before you commit to the design

Printed mechanisms are cheap to iterate and expensive to redesign, so the right move is to build the joint before you build the machine. Three tests, each one print and about ten minutes, settle most questions.

1. The clearance ladder. Print a small strip with five bores in a row at 0.15, 0.25, 0.35, 0.45 and 0.55 mm radial clearance, using the same orientation and material as the real part. Drop the real pin through each. The one that turns with a slight drag is your number, and it is not universal, it is your printer's number for that material and that orientation. Write it down. This single test eliminates most of the guesswork in every future mechanism.
2. The thirty-cycle warm-up. Any joint intended to move should be cycled thirty times by hand before it is trusted. If it gets looser, the material is wearing, and that joint needs a real bearing surface. If it gets tighter, the material is smearing and the contact pressure is too high. If it stays the same, the design is roughly right. Thirty cycles is enough to find the failure and short enough to do before assembly.
3. The hot-day test. Put the assembled mechanism somewhere at 45 to 50 °C for an hour, then try it cold, then warm. A mechanism that only binds when warm is a clearance problem caused by differential expansion, and it will surface in service rather than on the bench. A mechanism built in PLA will simply deform.

For anything that is meant to hold a position for months, the thirty-cycle test needs a companion: load it, leave it loaded for a week, and measure whether the position has changed. That is a creep test, and it is the test printed mechanisms skip and then get caught by.

Ten rules, ordered by how often they matter

1. Count the constraints before you draw anything. One joint should locate, the rest should comply.
2. Print bores vertically so they are round, and pins horizontally so they are strong.
3. Start at 0.35 mm radial clearance on a 6 mm pin, then run the clearance ladder on your own printer.
4. Lengthen the boss before you change the material. Bearing pressure is the variable that decides seizure.
5. A slotted hole in one joint is worth more than a tighter tolerance everywhere.
6. Never pair a fibre-filled material with itself as a sliding surface.
7. Grease is a symptom. Geometry is the cure. Use dry film only.
8. A toggle needs 2 to 5 degrees past centre and a frame that does not flex.
9. Check the four-bar condition before printing, not after.
10. Thirty cycles by hand, then a week under load. Both tests are cheap; skipping them is not.

The overall shape of the problem is worth restating, because it is counterintuitive. A printed mechanism is not a machine that has been made badly. It is a machine whose joints are soft, whose parts are slightly wrong, and whose material is a fraction as stiff as metal. The designs that work are the ones that accept all three facts and arrange the constraints so that the errors do not have to be absorbed by force. Clearance, slot, compliance, orientation, and a boss that is longer than you first thought: that is the whole toolkit. It is a short list, and it is enough for almost everything a desktop printer is asked to make move.

None of this is visible in CAD. In CAD the mechanism is a set of lines and cylinders that all agree with each other perfectly. On the plate, an 0.08 mm over-extension on the first layer, a seam placed on the inside of a bore, and a layer plane running the wrong way through an ear turn a design that was geometrically correct into one that binds. The rest of this article takes the four failure families in order and asks what each one has to do with the numbers printed on the spool.

Seizure: the joint that welds itself shut

Seizure is the failure that surprises people most, because the mechanism moves freely when it is first assembled and then quietly refuses to move two weeks later. The usual report is that it worked, then it went stiff, then it stopped. When the joint is opened, the bore is not worn. It is polished, slightly dull, and in bad cases faintly smeared, with matching marks on the pin.

Two mechanisms produce this. The first is pressure plus dwell. A printed polymer held under a constant contact pressure at an ambient temperature well below its glass transition still creeps, and the creep closes the clearance. Once the clearance is gone, the contact pressure rises, the creep accelerates, and the joint reaches a state where the surfaces have conformed perfectly to one another and there is nowhere for anything to slide. PLA is the worst offender here, and the reason is not that it is weak. It is that its glass transition is low enough that a warm room, a bit of sunlight, or a motor dissipating two watts into its mount puts the contact surface within range of softening.

The second mechanism is cold welding by diffusion, and it needs much less pressure than people expect. Two clean, smooth, compatible polymer surfaces held in intimate contact at room temperature will bond over days to weeks, because polymer chains at the interface are mobile enough to entangle given enough time and enough contact area. Perfect conformity is exactly what a printed pin in a printed bore does not have, and that surface roughness is what saves most joints. The joints that weld are the ones where the clearance has already closed and the surfaces have polished themselves flat against each other.

What to change, in order of effort

• Pick the pairing before the clearance. PLA against PLA is the worst common combination in desktop printing, and PETG against PETG is one of the best. Same-material joints are not automatically bad; soft-and-smooth against soft-and-smooth is bad.
• Take the pivot out of the warm zone. A printed pivot within 20 mm of a motor body, a heated bed frame or a hotend shroud is a different design problem from the same pivot on a cold frame. Move it, insulate it, or use a material whose glass transition is far above the local temperature.
• Add a sacrificial sleeve. A thin printed bushing of a dissimilar material inside the bore, replaced when it wears, converts a welded joint into a maintenance item. This is the cheapest structural change available and it costs one extra part.
• Give the contact somewhere to go. A groove, a counterbore, a relief chamfer at each end of the bore. When the polymer creeps, it creeps into the relief rather than into the clearance.

Debris: the gap that fills up with its own wear

The second failure family is genuinely a printing problem, and it is the one that separates printed mechanisms from machined ones. When a printed surface wears, it does not produce smooth, rounded debris. It produces angular, abrasive particles on the order of the layer height, because the surface itself is a stair of partly-fused polymer. Those particles land in the clearance between the pin and the bore, and a clearance that was 0.3 mm becomes 0.2 mm, then 0.1 mm, and the joint goes from free to stiff to jammed in a way that looks like seizure but is not.

The diagnostic is unambiguous. Open the joint and look at the bore. Seizure leaves a polished, conformed bore. Debris leaves a bore with grey dust in the groove, a bright worn band, and often a visible ring of packed powder at the open end where the debris was pushed out and could not escape. That ring is the tell.

The fix is not a bigger clearance, because a bigger clearance holds more debris and takes longer to fill but fills just the same. The fix is a gap that cannot trap. Three geometries work:

Geometry	What it does	Cost	When to use it
Relief groove at each end of the bore	Lets debris migrate out of the load zone instead of packing	One feature in CAD, no extra part	Every printed pivot, without exception
Chamfer 45 degrees by 0.5 mm at bore ends	Removes the sharp edge that scrapes the pin and starts the dust	Trivial	Any joint the user will cycle more than a hundred times
Open-ended pin with a loose fit at one end	Gives debris an exit path rather than a dead end	Requires the pin to be retained elsewhere	Higher-cycle joints, motor-driven linkages
Dissimilar soft sleeve	Wears instead of the real bore, and produces softer debris	One extra printed part	Where the joint cannot be opened for cleaning

The chamfer deserves its own sentence because it is the most underrated feature in printed mechanism work. The sharp edge at the end of a printed bore is a scraper. Every cycle, it shaves a small amount of material off the pin, and that material goes straight into the joint. A 45-degree chamfer of half a millimetre removes the scraper entirely, and it costs nothing in print time and nothing in CAD effort. If you add one feature to a printed mechanism, add this one.

Overconstraint: when two good joints make one bad mechanism

A pinned joint removes two translations. Two pinned joints on the same rigid link, with parallel axes, remove two rotations from the remaining freedom of the link. That is the arrangement everyone builds for a parallel arm, a folding stand, or a lid stay, and it is geometrically correct on paper. It is also, in printed plastic, almost always slightly wrong, because the two pin spacings on the link are not exactly the two pin spacings on the base.

Printing error is not random noise that averages out. It is a consistent bias on a given printer and material, and both parts get the same bias. A link printed at 79.9 mm centre-to-centre and a base printed at 80.1 mm centre-to-centre are 0.2 mm apart, and the two pins cannot both be in their holes without something deforming. The deformation is elastic, so the mechanism assembles, carries a permanent internal preload, and then binds at one end of its travel where the preload adds to the external load.

The symptom is distinctive and worth memorising: the mechanism is stiff at one extreme and loose at the other, and it is stiff in the same place every time. That is not a clearance problem. Opening the clearance will make it loose everywhere and stiff in the same place. The fix is to remove the redundancy.

1. Slot one of the two joints along the direction where the link has no work to do. This is the best fix because it is free, it keeps the position accuracy of the pin that matters, and it makes assembly trivial.
2. Make one joint a flexure instead of a pin. A short printed leaf at the second joint costs two millimetres of length and absorbs the error entirely, at the price of a small restoring force.
3. Drill or ream after printing only where the mechanism genuinely cannot tolerate play. This is the expensive option, and it is worth it only if you need both pins accurate to better than 0.1 mm.
4. Print both parts in the same orientation on the same plate. The bias is then identical rather than opposed, and the error largely cancels. This is a free fix and it is routinely overlooked.

The general principle is worth stating once because it covers more than pivots. Every printed mechanism should have exactly one set of features that fully determines the position of every part, and every other connection should be able to comply in at least one direction. Joints that are over-constrained do not fail loudly. They run hot, wear fast, feel wrong, and then break somewhere unrelated.

Load reversal, ears and the one direction layers cannot take

The fourth family is the only one that involves strength, and it is where most of the visible breakages in printed mechanisms come from. A pivot boss is usually an ear: a boss standing on a plate or a tab, holding a pin, with the load arriving at some angle to the plate. The ear fails at its root, and it fails as a delamination rather than a tensile fracture, because the root is where the layers are stacked and the bending moment is highest.

Two numbers govern this. The first is the moment at the root, which is the pin load multiplied by the distance from the pin axis to the plate face. Doubling the boss height doubles the moment at the root while the root section stays the same, which is why tall bosses tear off. The second is the section modulus of the root, which grows with the square of the width and only linearly with thickness when the load is bending the tab.

The practical consequence is that a pivot boss should be as short as the mechanical clearance allows, and as wide at the root as the envelope allows. Where the boss has to be tall, a pair of triangular gussets at the root is worth more than a thicker boss, and the reason is the same as in any bending part: moving material away from the neutral axis does more than adding material near it.

Bosses printed as a cylinder on a plate put every layer interface in the plane of the bending moment. Rotating the joint so that the pin axis is horizontal, and the boss is printed lying down, aligns the layers with the load and typically doubles the load the ear can carry before delaminating. This is not a small effect, and it is the reason printed arm designs that look odd in CAD are usually the ones that survive.

Design rules that survive the move from CAD to plate

Everything above reduces to a short list of decisions that should be made before a mechanism is modelled, because after it is modelled they are expensive to change.

Decision	Default that fails	Default that works	Why
Bore orientation	Horizontal, to avoid supports	Vertical, no support needed either	Vertical bores are round; horizontal bores are a stack of chords
Clearance basis	Nominal value from a forum	A printed ladder test on your own machine	Shrink and scallop are printer-specific and material-specific
Joint count	Every joint fully located	One located, the rest compliant or slotted	Printed dimensions cannot support full redundancy
Boss height	As tall as the part needs to clear	As short as clearance allows, gussets if not	Moment at the root scales linearly with height
Bore edge	Sharp	0.5 mm chamfer both ends	A sharp printed edge is a lathe tool, not a bearing edge
Material pairing	Same material, cheapest grade	Soft sleeve in a stiff structure, or PETG on PETG	Self-mated hard polymers weld and powder
Debris path	Enclosed clearance	Relief groove and open end	A trapped particle is a lap in a bore

One more rule does not fit the table because it is about sequence. Print the joint first, in isolation, and test it before you print the mechanism it belongs to. A 20 mm test coupon with the actual bore, the actual pin and the actual material tells you the clearance, the seizure behaviour and the debris behaviour in about eleven minutes, and it will save a redesign. Mechanisms fail as systems but they are diagnosed as joints, and it is far cheaper to diagnose a joint that is not attached to anything.

Closing

The parts discussed here are the ones that keep a machine from being a sculpture. Pulleys, pins, links, latches, clamps and stays are not decorative, they are the interface between a design and a load, and they are where printed part quality stops being a matter of layer lines and becomes a matter of whether the thing works in six months.

The common thread through all four failure families is that each of them is a mismatch between what a printed polymer can hold and what the design asked of it. Seizure is a load held too long. Debris is a gap that holds what it removes. Overconstraint is a position demanded too precisely. Ear failure is a moment carried across the worst possible plane. None of these needs a stronger material. All of them need a design that puts the load where the polymer is good at carrying it, and lets the rest go.

That is the quiet advantage of designing for print. A machined mechanism can be made to work by tightening tolerances and adding metal. A printed one has to be organised correctly from the start, which is more work at the beginning and produces a mechanism that is lighter, cheaper, repairable and made of parts you can reprint in an afternoon.