Every guide tells you a 3D printed part can be watertight. Almost none of them tell you what to do when the part has to hold pressure while somebody takes it apart to change a filter, and puts it back together in a cold garage, and then leaves it in the sun. That gap is what this guide is about.
Sealing is not the same problem as being watertight. Watertight means the wall itself lets nothing through. Sealing means two printed surfaces meet, and nothing passes between them — under pressure, across temperature swings, through dozens of assembly cycles, and while the plastic underneath slowly relaxes and moves. The wall is the easy half. The joint is where parts leak.
Why FDM Sealing Is a Different Problem From Machined Sealing
A machined aluminium flange seal works because the seal is the only variable. The faces are flat to a few microns, the material is stiff enough that bolt torque does not change the geometry, and the surface will still be the same next year. None of those three assumptions survive contact with an FDM part.
Printed faces carry layer ridges. Printed material creeps under sustained load, so a compressed gasket loses preload over weeks rather than over years. And the same wall that resists your fluid happily passes air along the layer lines, which is why a part can hold water and still fail a vacuum test. If you design grooves and gaskets the way a machinist would, you will build a joint that seals on the bench and weeps in service.
The Five Failure Modes, and Which Ones Are Reversible
Before choosing a geometry, know which mode you are trying to prevent. They have different fixes, and two of them cannot be fixed by tightening anything.
| Failure mode | What it looks like | Root cause | Fix class |
|---|---|---|---|
| Insufficient sealing stress | Slow uniform weeping, worse when hot | Too little compression, or the gasket never reached its sealing pressure | Geometry and torque |
| Squeeze loss through creep | Seals when assembled, leaks two to six weeks later | The printed material relaxes under sustained compression | Material and preload reserve |
| Extrusion or pinch damage | Sudden leak after assembly, visible torn gasket | Over-compression, sharp groove corners, gasket squeezed past the groove | Groove fill ratio |
| Face ridging and bypass | Spray at one location, fine bubbles along a line | Layer ridges give a continuous leakage path across the sealing face | Surface prep and orientation |
| Chemical attack or swelling | Gasket swells, softens or crumbles after weeks in contact | The elastomer is incompatible with the fluid | Material selection |
Only the first is genuinely reversible by tightening. If the part has been creeping for a month, more torque buys you another week and damages the gasket while it does. If the gasket has swollen, nothing you do to the hardware will recover it.
Groove Geometry: The Numbers That Actually Matter
An o-ring groove has three jobs: keep the ring located, compress it a controlled amount, and leave nowhere for the compressed rubber to go except into itself. Printed grooves fail when the second or third job is skipped.
Groove depth is what sets squeeze. For a radial or face seal with a standard nitrile ring, target 20 to 30 percent compression of the cord cross section. A 3 mm cord therefore wants a groove roughly 2.1 to 2.4 mm deep. A 1.5 mm cord wants 1.05 to 1.2 mm. Below 15 percent compression the rubber never develops enough contact stress to seal; above 35 percent it takes a permanent set on the first assembly.
Groove width is what sets fill. The ring needs somewhere to displace as it is squeezed, but not room to roll. Target a groove width about 1.25 to 1.35 times the cord diameter, which puts the fill ratio near 75 to 80 percent once compressed.
Three details do more damage than any dimension error:
- Groove corners must be radiused. A sharp internal corner acts as a stress riser in the rubber and a notch in your print. A 0.2 to 0.4 mm radius at the groove floor is enough, and it is small enough to print reliably with a 0.4 mm nozzle.
- Groove sidewalls must be vertical, not drafted. Draft is helpful on a moulding tool and harmful here, because it makes the groove narrower at the bottom and lets the ring ride up instead of sitting flat.
- The groove must be a closed loop with no through crossings. A groove interrupted by a screw hole, a wire exit or a lamination seam has just given the fluid a doorway on the wrong side of the seal. Route the groove around every penetration in the mating face.
Compression, Squeeze and Fill Ratio by Cord Size
The table below is a practical starting point for face seals in printed hardware, using standard metric nitrile cord and a groove machined into the printed part. It assumes the mating surface is also printed and reasonably flat.
| Cord diameter | Groove depth | Groove width | Compression | Typical service |
|---|---|---|---|---|
| 1.5 mm | 1.05 to 1.20 mm | 1.9 to 2.0 mm | 20 to 30 % | Small enclosures, sensor ports |
| 2.0 mm | 1.40 to 1.60 mm | 2.5 to 2.7 mm | 20 to 30 % | Instrument housings, pump covers |
| 3.0 mm | 2.10 to 2.40 mm | 3.8 to 4.0 mm | 20 to 30 % | General plumbing, tanks, sumps |
| 4.0 mm | 2.80 to 3.20 mm | 5.0 to 5.4 mm | 20 to 30 % | Large covers, low-pressure vessels |
Two adjustments are specific to printed parts. First, if the groove floor is printed, the top surface of the floor is not the nominal surface — it is a stack of layers with a slightly rounded crown. Add 0.1 mm to the groove depth to account for it. Second, if the opposing face is printed rather than a machined lid, add another 0.1 to 0.15 mm of depth, because the opposing face's first layer lands slightly proud.
Flat Gasket Joints Without an O-Ring
Lots of real parts use a flat sheet gasket rather than a moulded ring, either because the shape is not a simple loop or because the part has to be field-serviceable with scissors and a template. Flat gasket joints live or die on three things: gasket thickness, bolt spacing and face flatness.
A flat gasket seals by being compressed between two faces until its internal pressure exceeds the fluid pressure. Thin gaskets need less bolt load but tolerate less face error. Thick gaskets bridge face error but need much higher bolt load to reach the same contact stress, and printed plastic will creep under that load. For printed housings, a compressed fibre or rubber-cork sheet of 1.5 to 3 mm is usually the sweet spot.
Bolt spacing deserves more attention than it gets. Between any two bolts, the gasket tries to push the faces apart and the middle of the span lifts. The rule of thumb is to keep bolt spacing no more than 8 to 10 times the gasket thickness. A 2 mm gasket therefore wants bolts every 16 to 20 mm at most. Most printed enclosures with a leak have exactly this problem, and it is invisible until you look at gasket witness marks and see that the middle of every span never touched.
Face Flatness: The Number Nobody Checks
Machined flange specs chase fractions of a millimetre per hundred millimetres. Printed faces cannot reach that, and pretending otherwise leads to over-torquing. What printed parts can reach is roughly 0.1 to 0.2 mm of deviation across a 100 mm face, provided the face is oriented well and the part is not warped.
The practical approach is not to chase absolute flatness but to make the seal tolerant of it. Three tactics work:
- Design in a sacrificial crush zone. Leave a small raised bead — 0.3 to 0.5 mm tall and about 1 mm wide — on one sealing face, just inside the groove or just inside the gasket outline. It crushes first, sealing at low bolt load, and lets the rest of the face carry the structure.
- Never seal on a face printed against the build plate and one printed at the top. The plate-side face is squashed and smooth; the top face is crowned. Pair faces that came from the same orientation, or design an intentional 0.2 mm compression allowance.
- Keep the sealed area away from warping-prone geometry. A large flat face on a part with long straight corners will curl. Ribs, a smaller sealing footprint, or a chamfered perimeter all help more than any amount of bolt torque.
Surface finish matters here too. Layer ridges do not block a leak; they give it a highway. A face printed at 0.10 to 0.16 mm layer height with reasonably high extrusion width is smoother than one printed at 0.28 mm, and that difference shows up directly as lower leak rate. If the face has access, a light pass with 400 then 800 grit on a flat surface will do more for seal integrity than any change to the groove.
Material Choice for the Rigid Side — It Drives Creep
Most people spend their attention on the elastomer and almost none on the plastic holding it. That is backwards, because the plastic is the part that creeps, and creep is what eventually costs you the seal.
PLA is a poor sealing material for anything under sustained load. It creeps at room temperature, it softens well below the temperatures a pump or engine bay will reach, and it becomes brittle as it ages. It is fine for a sealed decorative enclosure that is bolted once and never touched. It is a bad choice for anything that cycles pressure or temperature.
A drink bottle shell is PETG's natural comfort zone: it is tougher than PLA, slightly less creep-prone, and cheap enough to iterate with. But PETG still relaxes under sustained compression, so it needs either a spring element or a contact pressure test after a month in service. Standard ABS and ASA hold up better over time and take higher service temperatures, and they are the sensible default for enclosures that will sit outdoors. If the joint sees real pressure, the glass- or carbon-filled grades are worth the price, because the fibre reinforcement is exactly what suppresses creep. A small sealed housing that cycles pressure and temperature is the case where a reinforced polyamide or a stiff filled PETG earns the money back.
| Printed material | Creep resistance | Heat limit | Suitable sealing duty |
|---|---|---|---|
| PLA | Low — measurable movement at room temperature | About 55 to 60 degrees C | Sealed, static, indoor, never opened |
| PETG | Moderate | About 70 to 75 degrees C | Bench-top liquid handling, moderate pressure |
| ABS / ASA | Moderate to good | About 95 to 100 degrees C | Outdoor enclosures, pump covers, hot fluids |
| Filled PETG, PC or PA | Good | Varies with base polymer | Cycling pressure, high preload, thin flanges |
Whatever you choose, the same two habits apply. Print the flange area with solid walls rather than sparse infill, because infill pockets collapse under bolt load and turn a flat flange into a wavy one. And dry the filament properly before printing a sealing part, because wet filament produces microvoids in the wall — and those voids are connected channels through the flange. If you want to be deliberate about this step, the drying and storage routines in the drying and storage guide cover the temperature and time tables per polymer.
Tightening Bolts Without Killing the Joint
Bolted joints in printed plastic fail in a specific and predictable way. The bolt is strong, the plastic is not, and every bit of torque over what the plastic can carry is converted into crushing, cracking or creep.
Print a proper washer face rather than letting a bolt head sit on a printed surface. A flat washer spread over a 2 mm thick printed boss is vastly better than the bolt head alone, because it distributes the load over several times the area and keeps the surface from being locally crushed into the flange.
Tighten in a cross pattern, in at least two stages — snug all bolts first, then bring them to final torque — so the gasket compresses evenly instead of extruding to one side. Finally, use a torque wrench on the low end and accept the number it gives you. For M3 bolts into a printed part with a 3 mm washer, a working range around 0.4 to 0.6 Nm is realistic for ABS-class plastic and dangerously high for PLA. Above that, you are not sealing harder, you are embedding the washer and reducing preload.
Sealing on a Print, Not Against It Yes — Orientation Decides Everything
The single most effective decision in the whole design is where the sealing face lands in the print. Layer lines have essentially zero tensile strength across them. A face that was printed as a wall in the XY plane is smooth and nearly non-porous; the same face produced by a stack of top layers is a narrow strip of bonding lines.
Put sealing faces where they can be printed as either the first layer against a smooth plate, or as a vertically oriented wall. Avoid making a sealing face out of top-surface infill, and avoid any arrangement where the leak path runs between two layer boundaries — the classic case is a sealing flange printed flat, where fluid can travel along the interlayer bond from the inside of the part to the outside without ever crossing a perimeter wall. If your sealing face must be printed flat, increase extrusion width, use a lower layer height and add at least one extra perimeter, and consider printing a light oil-resistant coating on the finished face.
Resin printing sidesteps much of this: an SLA or LCD part is one continuous cured volume with no interlayer bond to leak along, and its faces take detail — sharp grooves, small radii — far better than FDM. The trade is in the other direction: cured resin will absorb some liquids and swell, and its solvent resistance varies enormously between resin types and between cure levels. If the sealed part is going to touch fuels, solvents or caustic cleaners, the swelling behaviour in resin in solvents, oils and caustics is worth reading before you commit to a resin body with an FDM lid, or vice versa.
Hardware Sealing, or the Quiet Argument for Mixing Processes
There is a whole class of sealing problems where the honest answer is: do not print the sealing face. Print the housing, and let a standard o-ring or a flat gasket do the sealing. An injection-moulded nitrile ring costs cents, is available in hundreds of sizes, is formulated for the fluid it will meet, and is the subject of decades of published data. A printed groove plus a purchased o-ring is a better seal than a printed o-ring in almost every scenario, and it is far more repeatable.
A related and underused trick is to combine processes within one sealed assembly. An FDM body gives you inexpensive structure, tough exteriors and easy embedded threads. A resin or an FDM lid with a printed gasket channel gives you the fine groove detail. Where the assembly has to be opened repeatedly, put the fastener load into metal hardware rather than plastic threads, and read the failure modes in creep, fatigue and preload before you decide how hard to torque the cover.
Pressure, Vacuum and the Test That Proves a Seal
You cannot tell whether a seal works by looking at it, and the classic kitchen test of filling it with water and watching is only sufficient for one of the five failure modes above. Water is a relatively viscous, high-surface-tension fluid that a fine interlayer channel will hold back — the same channel will pass air and vapour straight through. That is why parts pass a water test and fail on air, and why the industry standard for leak tightness is measured in gas, not liquid.
A workable bench procedure for a printed part looks like this:
- Test with air, not water, first. Cap every opening except one, pressurise to about half the intended working pressure, and apply a soapy water film to the joint. Bubbles locate the leak precisely, which a wet exterior will not.
- Use vacuum when the part will see vacuum. Draw a partial vacuum, then hold it and watch whether the reading decays. Vacuum testing is stricter than pressure testing and will expose layer-line channels that a pressure test misses.
- Repeat after thermal cycling. Heat the assembly to its service temperature, hold it, cool it, then retest. Most creep-driven squeeze loss shows up here rather than in the first hour.
- Retest after a disassembly cycle. A gasket that sealed on first assembly and leaks on the third tells you the groove is allowing extrusion or setting. That is your fill-ratio number being wrong, not your torque.
Nine-Point Checklist Before You Print the Sealed Version
Run this list before committing to a long print. Every item here has cost somebody a rebuild.
- Orientation chosen with the sealing face in mind, not after the geometry was finished.
- Groove depth set for 20 to 30 percent squeeze at the actual cord size, plus 0.1 to 0.2 mm for printed faces.
- Groove width about 1.3 times the cord, with radiused floor corners and vertical sidewalls.
- Groove loop closed around every bolt hole, cable entry and lamination seam.
- Bolt spacing no more than 8 to 10 times gasket thickness for flat gaskets.
- Washer faces designed in, with bolt torque in the 0.4 to 0.6 Nm range for M3 into ABS-class plastic.
- Flange region printed solid, with no sparse infill under the bolt load path.
- Filament dried before the sealing part is printed, so the wall has no connected voids.
- Leak test planned with air or vacuum, and repeated after thermal cycling and after reassembly.
None of this requires exotic hardware. It requires deciding, before the first line goes down, that the joint is a designed feature rather than a gap that will be filled with sealant later. Sealed parts that hold pressure for years are not the ones with the most torque on the bolts. They are the ones where somebody worked out where the leak would go, and then closed every route to it.
Building something that has to stay dry under pressure? The ANTINSKY range covers the rigid side of the problem — PETG and ABS for flanges, ASA where the sealed housing lives outdoors, flame-retardant ABS for electrical enclosures, and filled polyamide where stress and load have to share the same part. Add a filament dryer to keep the wall solid, and a tungsten-carbide nozzle to print it.