Troubleshooting guide

Metal 3D Printing Defects: Symptoms, Causes, and Fixes

A shop-floor guide to catching and correcting metal 3D printing defects before a build is scrapped. Written for engineers and buyers who need to decide whether to re-slice, re-fixture, or move the part to CNC.

Get Instant Quote Talk to an Engineer

12-hour quote±0.005 mm CNC fallbackNo MOQ

Metal 3D printing defects on a laser powder bed fusion build

On this page

1
Symptom, likely cause, and first action
2
What metal 3D printing defects actually look like
3
Scan parameters that keep the melt pool stable
4
Orientation choices that prevent defects before the first layer
5
When metal 3D printing defects mean you should machine instead
6
Step by step: from symptom to corrected build

Symptom map

Symptom, likely cause, and first action

Read down the first column until the symptom matches what you see on the plate or in the scan data.

Symptom	Likely cause	First action
Layers not bonded, part lifts off plate	Residual stress, weak supports	Re-check support density and preheat
Porosity in CT or cross-section	Keyhole mode, gas entrapment	Cut laser power, raise scan speed
Dimensional drift over 150 mm	Thermal shrinkage buildup	Re-scale model, add stress relief
Rough top surface on downskins	Poor powder spreading	Check recoater blade and powder
Cracks at sharp internal corners	Fast cooling, stress riser	Add fillets, change scan strategy
Melt pool glow drifts during build	Laser power or focus drift	Pause, log the layer, inspect optics
Delamination after heat treat	Trapped stress released late	Stress relief before HIP or aging

Failure modes

What metal 3D printing defects actually look like

Most metal 3D printing defects start as a thermal event, not a mechanical one. The laser melts powder, the melt pool solidifies in microseconds, and the part cools faster than the surrounding powder bed. That mismatch builds residual stress layer by layer. By the time you see a warp, the stress has already been locked in for hundreds of layers.

Some defects are visible on the plate: a lifted corner, a rough downskin, a crack running from an internal corner. Others only show up in CT or after sectioning. Porosity under 100 μm is easy to miss on a visual check and still fails a fatigue spec. That is why in-process monitoring matters more for metal than for polymer printing.

Not every defect is worth fixing in the printer. If the geometry is a bracket, a housing, or any part with simple internal features, 5-axis CNC machining often reaches the tolerance faster and with a known surface finish. The printing route makes sense when internal channels or lattice features cannot be machined at all.

One more thing worth saying plainly: a single bad build rarely has a single cause. A rough top surface plus a lifted edge usually means the powder layer was inconsistent and the stress was already high. Fix the layer first, then revisit the scan strategy.

1
Visible on the plateWarp, cracks, rough downskins, short powder spreads.
2
Only in inspectionGas porosity, lack of fusion, internal voids.
3
Only after heat treatDelayed cracking, dimensional shift, hardness scatter.

Process window

Scan parameters that keep the melt pool stable

Laser powder bed fusion runs on a narrow window. Too much energy density and the melt pool goes into keyhole mode, drilling a deep vapor cavity that collapses and traps gas. Too little and you get lack of fusion: powder particles that never fully melt, leaving elongated voids between layers. Both fail a CT scan, and both look different under a microscope.

Energy density is the number most teams track first. For Ti-6Al-4V on a 400 W laser, a workable band sits near 55 to 75 J/mm³ with a 60 to 100 μm layer. Below that band, lack of fusion dominates. Above it, keyhole porosity and spatter rise quickly. The exact numbers move with spot size and scan strategy, so treat them as a starting point, not a recipe.

Preheat changes the picture. A bed heated to 200 °C reduces the thermal gradient and lowers cracking risk in tool steels. For some alloys, a heated build plate to 400 °C or higher reduces stress enough that thin walls survive without heavy support. The trade-off is longer heat-up and cool-down, plus powder that behaves differently when it is warm.

In-process monitoring is what turns this from guesswork into data. A camera watching the melt pool can flag a layer where the glow drifts outside the normal band. That does not fix the defect by itself. It tells you which layer to inspect, so you can decide whether to stop the build or let it finish and measure.

1
Too much energyKeyhole porosity, spatter, rough top surfaces.
2
Too little energyLack of fusion, weak layer bonding.
3
Preheat above 200 °CLower gradients, less cracking in tool steels.

Geometry and orientation

Orientation choices that prevent defects before the first layer

Orientation decides which faces need support and where heat will accumulate. A part laid flat on the plate has good heat conduction through the base but needs supports under every overhang. A part tilted 30 to 45 degrees spreads the overhang angle so supports are shorter and easier to remove, at the cost of a taller build and more layers.

Sharp internal corners are a common defect site. The laser turns a tight corner quickly, the melt pool overheats locally, and the next layer sits on a rough surface. Adding a 0.5 to 1 mm fillet spreads the heat and usually removes the crack. It is a small CAD change that saves a whole build.

Thin walls below 0.4 mm are hard to print reliably in most alloys. The wall may survive the build and then distort during support removal. If the design needs a 0.3 mm wall, expect a high scrap rate and plan for machining the wall to final thickness instead of printing it to size.

Support removal is where a good build often becomes a bad part. Cutting supports off by hand puts load into thin sections. For anything with a tight tolerance callout, we plan the machining sequence so the critical faces are cut after supports come off, not before.

1
Tilt 30–45°Shorter supports, fewer heat traps.
2
Fillet internal corners0.5–1 mm fillet spreads local heat.
3
Walls under 0.4 mmHigh distortion risk during support removal.

When to switch

When metal 3D printing defects mean you should machine instead

Some parts are simply a bad fit for additive. A part with no internal channels, no lattice, and no conformal cooling has little to gain from printing. If the same geometry can be cut from 6061 or 17-4PH bar stock on a 5-axis machine, the route is shorter and the tolerance is predictable.

The decision usually comes down to feature access. If the critical surfaces are reachable with a 3 mm tool, CNC can hold ±0.005 mm and a finish between Ra 0.8 and 1.6 μm. If the feature is a curved internal channel with a 4 mm diameter, no end mill reaches it and printing stays the only option.

There is also a hybrid path. Print the blank with extra stock on the critical faces, stress relieve it, then machine those faces to size. That works well for impellers and manifolds where the internal geometry must be printed but the mounting faces and bores must be precise. The printed surface is never the final datum.

For buyers comparing quotes, ask one question: which faces carry the tolerance callout? If the answer is a machined face, the printing step is a blank-making step, and the schedule should be planned around machining, not around the printer.

1
Machine itNo internal channels, all faces tool-reachable.
2
Print itConformal cooling, lattices, curved internal channels.
3
HybridPrint the blank, machine the critical faces after stress relief.

Workflow

Step by step: from symptom to corrected build

Run these in order. Fixing the layer before touching the scan strategy saves a build.

1
Record the symptom on the platePhotograph the lifted edge or rough skin and note the layer number. Pull the build log for laser power, scan speed, and bed temperature at that layer.
2
Check the powder layer firstMeasure spread consistency across the plate. Look for streaks or bare patches. Replace the recoater blade if wear exceeds the supplier limit, and sieve powder to remove agglomerates.
3
Recompute energy densityTake the logged power, scan speed, hatch distance, and layer thickness. Compare against the alloy band, for example 55–75 J/mm³ for Ti-6Al-4V. Move in small steps, not large ones.
4
Adjust preheat and stress reliefRaise bed temperature in 50 °C steps within the machine limit. For tool steel, plan stress relief before any hardening step so residual stress releases in a controlled cycle.
5
Change orientation or supportsTilt the part to 30–45° if overhangs dominate. Add 0.5–1 mm fillets at internal corners. Increase support density only where heat accumulates, not everywhere.
6
Re-scan the model for shrinkageScale the model by the measured deviation from the last build. If a 150 mm span drifted 0.4 mm, correct the scale before the next run rather than after.
7
Inspect the first article fullyCT or section the first part off a corrected build. Check porosity size and location against the drawing. Do not release the run until the first article passes.

FAQs

Common questions

Can a defect be fixed after the build instead of scrapping it?

Sometimes. Surface porosity open to the outside can be removed by machining 0.2 to 0.5 mm off the skin, then re-inspected. Internal voids cannot be machined away, so a part with internal porosity usually gets scrapped or downgraded to a non-critical use.

Cracks are worse than porosity. A crack at an internal corner often continues under load even after the visible tip is ground out. If the part carries a fatigue spec, we treat a crack as a reject.

How do I know if porosity is from the powder or from the laser?

Look at the void shape. Gas-entrapped porosity from the powder or keyhole mode tends to be round and small, often under 100 μm. Lack of fusion voids are irregular and elongated, and they follow layer boundaries.

Powder contamination shows up as voids clustered near the surface or near support contacts. Laser-driven keyhole porosity is more evenly distributed through the melt track. The two need different fixes, so the shape matters.

Does heat treatment always cause cracking?

No, but it releases whatever stress the build stored. If the build already has a crack tip, heat treatment can open it. If the part was stress relieved in the printer or in a furnace right after the build, the risk drops sharply.

The common mistake is skipping stress relief on a thin-wall part because the build looked fine. The part measures well on the plate and moves 0.3 mm after aging.

What tolerance should I expect from a printed metal part?

As-built metal printing typically lands in the ±0.1 to ±0.3 mm range on a 100 mm part, and it varies with alloy and orientation. That is far from a machined tolerance.

Where a drawing calls for ±0.005 mm, the printed surface cannot be the final surface. Plan a machining pass on those faces after stress relief.

Can you machine a printed blank without damaging the internal features?

Yes, if the setup is planned. We hold the blank on sacrificial stock, machine the datums first, then cut the critical faces in the same setup where possible. Internal channels stay untouched because the tool never enters them.

This is the usual route for impellers, manifolds, and conformal-cooled inserts. The printed geometry supplies the internal shape; the machined faces supply the fit.

How fast can a corrected build or a CNC fallback ship?

Quotation and DFM feedback come back within 12 hours, and production can start within 24 hours of approval. Standard machined parts ship in 3 to 5 days.

For printed blanks that then get machined, add the build and stress relief time to the schedule before the machining window starts.

Send the drawing, get a route back

Upload the part and the inspection report if you have one. We will tell you whether to re-slice, re-orient, or machine the critical faces on a 5-axis center.

12-hour quote100% inspection before shipmentNo minimum order quantity

Get Instant Quote Talk to an Engineer