Why printed parts are anisotropic
A printed part is a laminate. Each line of plastic is well bonded to its neighbor in the same layer, but the bond between layers is formed by plastic that has cooled while the next layer was being laid, so it is weaker. The strength therefore depends on the direction of the load relative to the layers:
| Load direction | What fails | Typical strength relative to the best direction |
|---|---|---|
| Along the lines (X–Y plane) | The plastic lines themselves | 100 % (reference); the strongest direction |
| Across the layers in tension (Z) | Layer-to-layer bonds | Often 30 to 70 % for PLA and PETG; as low as 20 to 40 % for badly printed or cold-printed parts; better for hot, slow, enclosed prints of ABS, ASA and nylon |
| In compression | Buckling and local crushing | Usually 80 to 100 % of tension for solid parts, with the layers and walls staying aligned |
| In bending | Outer surface layers | Depends on which face is in tension: layers across the tension face are weakest |
| In shear between layers | Layer bonds | Weak: avoid shear loads that slide layers over each other |
A datasheet value for the plastic is the starting point, not the answer: a printed part is lower and directional. Design rules follow: align the load with the layers, avoid tension across the layers, print hot enough for good bonding, and add a safety factor of 2 to 4 over a conservative design stress (see below).
Worked example: orientation knock-down
A PETG filament has a printed tensile strength of 45 MPa along the lines. If the layer bond is 55 % as strong, a hook printed standing up breaks at about 45 × 0.55 = 24.8 MPa, and the stress at the sharp inside corner of the hook is raised by a stress concentration factor of 2 to 3 (see Fatigue and Stress Concentration), so it fails at a nominal stress of 8 to 12 MPa: less than a third of the datasheet strength. Printed lying down, with a fillet in the corner, the same hook carries roughly three times as much.
Walls, infill and what carries the load
In bending and torsion the stress is highest at the surface, so the outer shell (the walls) carries the load and the infill only supports it and keeps the walls from buckling. Adding a wall adds stiffness where it counts; adding infill adds mass where the stress is low. The chart shows a 40 × 20 mm bar in bending: at the same mass, more walls give clearly more stiffness than the same plastic spent on infill.
- A bar with 4 walls and no infill has a relative stiffness of 0.38 at a mass of 0.25. A bar with 1 wall and 25 % infill is heavier (0.30) yet less stiff (0.33): the walls give 1.4 times the stiffness per unit of mass.
- Doubling infill from 15 % to 40 % with 2 walls raises the stiffness only from 0.33 to 0.52 (61 %) at a mass increase of 83 %.
- The model treats infill as a core of reduced density. Real infill patterns carry shear and compression well but add little in tension, so these are estimates: print and test.
Estimating the strength of a printed part
Worked example: a cantilever arm
- A PETG arm is 20 mm wide and 10 mm deep, 100 mm long, printed lying flat with 4 walls (shell 1.80 mm thick). The design stress is 20 MPa: about half of the printed strength, to allow for stress concentrations, creep and a safety factor.
- Section. I of the whole section is 1 667 mm⁴; the hollow core (no infill) has 358 mm⁴; the shell therefore has I = 1 308 mm⁴ and S = I ÷ (H ÷ 2) = 262 mm³, against 333 mm³ for a solid bar: 79 % of the solid section’s strength with only 48 % of the cross-section filled with plastic.
- Capacity. M = σ S = 20 × 262 = 5 234 N·mm, so the tip load is P = M ÷ L = 5 234 ÷ 100 = 52 N (about 5.3 kg). A solid bar would carry 67 N.
- Check: test a printed sample at twice the working load before relying on it. The beam formulas and section properties are in Beam Bending and Deflection and Section Properties and Structural Profiles.
- Design stress rule of thumb: 20 to 30 % of the datasheet tensile strength for steady loads along the layers; 10 % or less for across the layers or cyclic loads.
- Stiffness: printed elastic modulus is 60 to 90 % of the datasheet value along the lines, lower across the layers, so deflection is larger than a solid-plastic calculation suggests.
- Stress concentrations: sharp corners, holes and thin sections break printed parts: use fillets and generous radii.
- Fatigue and impact: printed parts are notch sensitive and fatigue at low stresses; use a safe design stress and check any part loaded repeatedly (see Fatigue and Stress Concentration).
Heat, creep and long-term loads
Plastics lose stiffness and strength as they warm, and under a steady load they creep: they keep deforming slowly. PLA starts to soften near 55 to 60 °C, PETG near 70 to 80 °C, ABS and ASA near 95 to 100 °C, nylon and PC at higher temperatures. A part under a steady load needs a much lower design stress than a short-term load: a rule of thumb is 25 % of the short-term strength for a load held for months, and less at higher temperature. A bracket that holds a weight for a day may sag after a month.
- Do not use PLA in a hot car, near a heater or in direct summer sun; it deforms at 60 °C.
- Avoid steady tension and bolt preload on thin printed parts: they relax. Use inserts, washers and metal reinforcements, and re-tighten.
- Annealing (heating a part to just above its glass transition for an hour or so) can raise the heat resistance of PLA, PETG and nylon, but the parts shrink and distort (see Post-processing and Finishing).
- Chemicals and water: nylon absorbs water and softens; PLA hydrolyses slowly in warm, wet conditions; PETG and ABS handle many chemicals, but check each one.
Dimensional accuracy and shrinkage
A well-tuned FDM printer holds about ±0.1 to ±0.3 mm on small features, and worse on large parts. Plastics shrink as they cool: PLA about 0.2 to 0.3 percent, PETG 0.3 to 0.5 percent, ABS 0.4 to 0.7 percent, nylon 1 to 2 percent (typical figures), which adds up on big parts and is why warping starts at the corners. Typical errors come from: shrinkage, steps per millimeter, belt tension, loose or worn parts, nozzle wear, thermal expansion of the frame, elephant foot on the bottom layers, and the compensation of the slicer.
| Material | Shrinkage (typical) | Error on a 100 mm part | Scale factor to compensate |
|---|---|---|---|
| PLA | 0.3 % | 0.30 mm | 100.30 % |
| PETG | 0.4 % | 0.40 mm | 100.40 % |
| ABS / ASA | 0.6 % | 0.60 mm | 100.60 % |
| Nylon | 1.5 % | 1.50 mm | 101.52 % |
To compensate, print a calibration cube or a long bar, measure it with calipers, and scale the model in the slicer by the ratio of the target to the measured size: for a 100 mm cube that comes out 99.7 mm, scale by 100 ÷ 99.7 = 100.30 %. Compensate each axis separately if they differ. Do not compensate for features that depend on the nozzle (holes and thin walls): scale only for shrinkage.
How printing compares with machining
The international tolerance grades (ISO 286, IT) give the width of the tolerance zone for a size range. For sizes from 18 to 30 mm:
| Grade | Tolerance zone (18–30 mm) | As a plus-minus value |
|---|---|---|
| IT7 | 21 µm | ±0.011 mm |
| IT9 | 52 µm | ±0.026 mm |
| IT11 | 130 µm | ±0.065 mm |
| IT12 | 210 µm | ±0.105 mm |
| IT13 | 330 µm | ±0.165 mm |
| IT14 | 520 µm | ±0.260 mm |
| IT15 | 840 µm | ±0.420 mm |
| Process | Typical tolerance grade | Typical surface roughness Ra |
|---|---|---|
| FDM | IT13 to IT15 (±0.15 to ±0.4 mm on 20 mm) | 10 to 25 µm (layer lines) |
| MSLA resin | IT11 to IT13 (±0.07 to ±0.15 mm) | 1 to 5 µm |
| SLS / MJF | IT12 to IT13 (±0.1 to ±0.15 mm) | 6 to 12 µm |
| CNC milling and turning | IT7 to IT9 (±0.01 to ±0.03 mm) | 0.4 to 3.2 µm |
The takeaway: a printed part is a few IT grades worse than a machined one, so tight fits, bearing seats and sealing surfaces are best finished by drilling, reaming or machining after printing, with the printed part left oversize. For how fits and grades are specified see Fits and Tolerances and GD&T.
Clearances and fits
| Fit | Clearance per side (typical) | Use |
|---|---|---|
| Press fit | −0.05 to −0.15 mm (interference) | Pins and inserts that should hold without glue; chamfer the lead-in |
| Snug / locating fit | 0 to +0.1 mm | Parts that locate with a light push |
| Sliding fit | +0.15 to +0.25 mm | Parts that must move freely |
| Loose or print-in-place | +0.3 to +0.5 mm | Hinges, gears and moving parts printed assembled |
- Print a test piece (a ladder of holes or a stepped plug and socket) with different clearances on your own printer, material and settings, then use the best one. Fits drift with temperature, filament batch, nozzle wear and slicer settings.
- Orientation matters: a vertical hole is more circular than a horizontal one; fits across the layers behave differently from fits along them.
- Use the 3D Printing Tolerancing Guide to compute clearances for your tolerance and printer accuracy.
- Allow for thermal and moisture change: nylon swells when wet; a nylon bearing fit that is perfect dry can bind in a humid shop.
Testing and verifying
- Print test coupons beside the real part with the same settings and loads, and break them. Even a hand test (bending a printed strip until it breaks) shows the layer direction and the quality of the bond.
- Proof-load the part to twice (or more than) the working load, hold it for the time the load is applied (creep), and look for permanent deformation or cracks.
- Measure critical dimensions with calipers or a micrometer, at room temperature and after cooling, and record them for each print run.
- Never rely on a printed part as the only protection for a safety-critical load; use metal, a redundant path or a rated component.
Related
- 3D Printing and Additive Manufacturing: the overview of this guide.
- Designing for 3D Printing: orientation, snap fits and holes.
- Filament Materials: heat resistance and properties of each filament.
- Structures and Strength of Materials: the stress, beam and fatigue methods used above.
- Fits and Tolerances and 3D Printing Tolerancing Guide.
FAQ
How strong is a 3D-printed part?
It depends on the plastic, the orientation, the number of walls and the temperature. Printed parts are usually weaker than molded ones, and much weaker across the layers than along them. More walls and the right orientation help far more than more infill.
What infill should I use for strength?
15 to 25 percent of gyroid or cubic is enough for most parts; raise the number of walls to 3 to 5 instead of raising the infill.
How accurate is a 3D printer?
A well-tuned FDM printer holds ±0.1 to ±0.3 mm on small features; resin printers hold better. Shrinkage dominates on large parts, and holes come out undersize.
Can I use a printed part under a bolt load?
Yes, with inserts or a metal sleeve, washers to spread the load, and a modest preload. Plastic creeps, so re-check the preload, and never rely on a printed thread for a load that could injure someone.
References
- Shigley, J. E. and Mischke, C. R., Mechanical Engineering Design, McGraw-Hill (beam stress and safety factors).
- ISO 286-1, Geometrical product specifications (GPS): ISO code system for tolerances on linear sizes (IT grades).
- Ahn, S. et al., “Anisotropic material properties of fused deposition modeling ABS”, Rapid Prototyping Journal, 2002; filament makers’ data sheets (printed strength by orientation).
- Machinery’s Handbook (tolerance grades, thermal expansion and shrinkage).
Last reviewed: October 2026. Strengths, knock-down factors and shrinkage are typical: they vary with material, machine and settings, so test your own parts.