← analyzer · failure modes · delamination

Layer separation in LPBF, and the cross-section jump behind it

Delamination — layers that separate from each other, or a part that splits along a build-direction plane — reads like a process fault, and often it is one. But there is a geometric cause that shows up repeatedly and is visible in the CAD file long before anyone slices it: the cross-sectional area changing too fast from one layer to the next.

The same predictor also drives balling and bead-up. Both are on the short list of failure modes that geometry can genuinely call.

Why a change in area is a change in risk

The laser deposits energy into whatever the current layer is. The heat leaves through the material below. When the area of a layer changes sharply, that balance breaks in one of two directions:

Area growing quickly going up means new material is starting over loose powder rather than over solid metal. Powder is a thermal insulator by comparison, so the new region runs hot, contracts hard, and pulls against the solid it is anchored to. This is the unsupported shoulder, and it is the classic corner-lift and recoater-collision site.
Area shrinking quickly going up funnels the heat from the larger section below into a smaller escape path. The narrowed section runs hot for the same reason a bottleneck does, and thermal gradient across a bond line is what pulls layers apart.

Growth is the one that leads, because it is the one that lifts material off the plate. Shrink is reported too, because a heat funnel is a real risk even when nothing lifts.

What the screen computes

The section_jump gate slices the part and walks the layer profile, taking the fractional area change per millimetre of Z between consecutive layers:

rel = (A₁ − A₀) / A₀ / pitch

Normalizing per millimetre matters: it means the number keeps the same meaning when the voxel pitch changes, so a value from a coarse screen and a value from a fine one are comparable. The report gives the worst growth and the worst shrink, each with the areas either side and the Z height where it happens, so you can go find it on the model rather than hunt for it.

Gaps in the layer profile are skipped rather than treated as jumps — a discontinuity in the slice stack is not a section change, and counting it as one manufactures findings.

What to do about one

A flagged jump is not automatically a redesign. In rough order of cost:

Reorient. The same part rotated puts the area transition somewhere else entirely, and often removes it. This is the cheapest fix and the one most often skipped, because trying it by hand means re-exporting and re-uploading.
Chamfer or fillet the transition so the area ramps over several millimetres instead of one or two layers.
Support the shoulder if the growth has to stay — which turns a build risk into a removal-labour cost, a trade worth making knowingly.

The geometry is a contributor, not the whole cause. Lack-of-fusion porosity and solidification cracking are also implicated in layer separation, and for both of them the geometry is the weak term. Energy density, powder-bed quality and thermal history dominate. The screen labels those modes “weak” in its own coverage table rather than claiming them. A clean section_jump result does not mean your layers will bond.

Scope, stated once. This is a necessary-condition checker, not a build-outcome predictor. Geometry can say a part is at high risk. It cannot say a build will succeed — that also depends on powder, machine, layout, parameters and the operator. Of the 18 LPBF failure modes in the tool’s published taxonomy, 4 are fully geometry-predictable and 12 are informed by at least one of the 20 gates; the rest are out of reach of geometry and are listed as such in every report.

Check a part

Every gate named on this page runs automatically on an uploaded STL or STEP at amprintanalyzer.com — free, no sign-up, and the geometry is never modified. Findings are pinned to the model and each one is labeled by its evidence class: measured (a direct geometric fact), screening (a dimensionless index that localizes risk), or estimate (a planning number with its assumptions stated).

If the screen disagrees with something you have actually measured on a printed part, that is the useful case — say so through the feedback strip. Disagreements are the calibration data.