How to Diagnose a 3D Print Failure From a Photo: A Practitioner’s Decision Tree

To diagnose a 3D print failure from a photo, you read three things in order: what the defect looks like, where on the Z axis it happens, and which of four physical categories it belongs to — thermal, mechanical, extrusion or digital. Get those three right and you’ve usually narrowed six possible causes down to one before you touch a single slicer setting. Most FDM failures trace back to the same short list: poor bed adhesion, wrong temperature, wet filament, a nozzle clog, a mechanical motion fault, or a mismatched slicer profile. The photo tells you which.

The discipline that separates a fast fix from a week of wasted filament is simple: match the failure by appearance first, follow the cause-and-fix chain, and change one variable at a time. The single most common mistake is changing three settings at once, printing again, and having no idea which one mattered. This guide gives you the visual signatures and the decision logic to work from a photo — the same reasoning our Vision-AI Diagnose tool applies when it reads an image.

The one question a photo answers instantly: at what height did it fail?

Before you name the defect, look at the failure height. This is the highest-value diagnostic shortcut a photo gives you, and it splits the entire problem space in two:

• A failure that happens at the same Z height every time is mechanical. No slicer setting will fix it. A layer shift, a Z-axis bind, or a lead-screw fault repeats at a consistent point in travel because the machine is failing at a physical location, not at a value in your profile.
• A failure that appears gradually or scales with geometry is thermal or extrusion-related. Warping worsens as the part gets larger; stringing appears wherever the toolpath travels; under-extrusion tracks with flow and temperature.

So the first pass on any photo is a category triage. Ask which of these four buckets the visual evidence points to:

If a print matches more than one category — and long prints often do — start with the most visually dominant defect. And before you go deep on any slicer parameter, check the three that cause most failures: Z offset, wet filament, and bed temperature.

Reading the visual signatures

Below are the defects you can identify from a clear photo, the physics behind each, and the fix chain. The key skill is telling apart failures that look similar but have opposite causes.

Under-extrusion vs over-extrusion

Under-extrusion shows as little gaps in the sides of walls or supports, missing layers, thin or dotty extrusion, and small holes in otherwise solid areas. The part comes out spongy, weak and noticeably light. The instinct is to crank nozzle temperature to force more plastic through — don’t. If the printer is under-extruding, the cause is far more often a clog, path resistance or a flow-rate limit than a genuinely low temperature. On Bowden setups, check the tube for debris, internal wear that raises friction, or too tight a bend radius; and confirm the tube is seated fully into the hotend, because any gap lets melted filament leak around it and clog.

Over-extrusion is the opposite signature: bulging walls, messy surfaces, blobs, oozing and poor dimensional accuracy, sometimes with layer drooping or jams. The fix chain is reduce the extrusion multiplier, calibrate E-steps, then lower temperature.

Stringing and oozing

Thin, hair-like threads bridging the gaps between separate features are stringing — small amounts of still-liquid filament on the nozzle dragged into a thread as the head moves from where it stopped to where it restarts. Oozing is the broader term for any extraneous leaking; stringing is specifically those deposited hairs. It’s a near-universal FDM issue, worst with PLA, PETG and TPU. Causes: wet filament, nozzle temperature too high, untuned retraction, or a naturally oozy material like PETG. Fix by tuning retraction, enabling wipe/coast, and lowering nozzle temperature 5°C at a time until it clears — but not so far that you tip into under-extrusion.

Warping vs delamination

These both come from uneven cooling, but they present differently. Warping is lifted corners and deformed edges: the plastic cools unevenly, shrinks, and peels off the plate. Delamination (layer separation) is a clean horizontal crack up the side of the part, often with a slight upward bend at the split. It happens when internal contraction forces exceed the layer-to-layer adhesion strength and the layers pull apart.

The distinction matters because the fix differs. Warping is an adhesion-versus-shrinkage problem: every warping failure traces back to uneven cooling overwhelming first-layer adhesion, not to one wrong number. PLA hides it until parts get large; PETG exposes it at edges and corners; ABS makes it nearly unavoidable without a controlled environment, because ABS contracts far more than PLA or PETG as it drops from extrusion temperature to ambient, loading strong internal stress into the base layers.

Delamination, by contrast, is a bonding problem. With very good bed adhesion — say ABS slurry on a large open-air part — you can get delamination instead of warping: the base stays stuck, so the stress relieves by splitting a layer higher up. Raise the hotend temperature 5–10°C at a time to improve bonding. For ABS, kill part cooling entirely — cold air on a hot ABS layer is the classic trigger — and print in an enclosed, ideally heated chamber. For PETG, run part cooling at 30–50% or off for the first few layers to strengthen the base bond.

Elephant’s foot

The bottom few layers bulge wider than the rest of the part. The first layer is squished too far: the nozzle sits too close to the plate and exerts pressure that spreads the extrusion — a z-stepping effect — often compounded by too-high bed temperature or excessive first-layer flow. Fix by nudging bed temperature down slightly and raising the Z-offset in small increments. If you’re wrestling this on a Bambu machine, our guides on first-layer adhesion on Bambu Lab printers and the exact P1S settings cover the trade-offs in detail.

Ghosting vs Z-banding — the pair people confuse most

Both show as surface texture, but they’re mechanically opposite and the photo tells them apart:

• Ghosting (ringing) is wavy — echo ripples that trail off after corners and sharp details, fading as they go. It’s a vibration/resonance problem from high acceleration, loose frame hardware or excessive speed. Fix by reducing acceleration and jerk and tightening frame bolts.
• Z-banding is straight horizontal ridges — a ribbed surface that repeats at a consistent vertical spacing up the whole wall. If the bands repeat at a fixed interval, suspect the lead screws, bent rods, couplers, dirty threads, or a machine binding at a regular point in its travel. Fix by aligning the lead screw, inspecting couplers and stabilising the Z frame.

One more check: if wall texture changes with flow, temperature or filament condition rather than repeating at a fixed interval, you’re looking at under-extrusion or a partial clog, not Z-banding. For layer-line splitting specifically, our piece on why prints split along layer lines goes deeper.

Layer shift

A sudden sideways offset at one layer — everything above shifted a few millimetres in X or Y from everything below, giving a stair-stepped look. This is almost always mechanical, so fix hardware before touching settings. Check belt tension (too loose skips steps, too tight binds — aim for a satisfying twang when plucked, with no more than a few millimetres of deflection under finger pressure), grub screws on pulleys, axis binding from debris or a snagged cable, and crashes where the nozzle clips a curled edge or a support and loses steps. Because it recurs at a fixed height when the cause is a physical obstruction, layer shift is a textbook case of the “same-Z-height = mechanical” rule.

First-layer problems and spaghetti

Uneven extrusion, gaps and peeling corners in the base point to a bad first layer: bed not level, nozzle too high or low, or wrong temperature. The first layer is the foundation — if it’s high, low, dirty, uneven or poorly attached, the rest fails even with a perfect model and profile.

Finally, spaghetti — chaotic strands of extruded plastic drawn in mid-air — is a symptom, not a root cause. Because an FDM printer runs open-loop, it executes G-code without checking whether plastic is landing correctly; once the model detaches from the plate, a support fails, or the nozzle clogs, the head keeps drawing toolpaths into thin air and the plastic cools into strands. When you diagnose spaghetti from a photo, you’re really diagnosing what failed first: adhesion, a support, or a clog. Catching that upstream is exactly what a G-code pre-flight check is for.

The limits of a photo — and how to work around them

A single photo is powerful but not omniscient. It can’t tell you filament moisture directly, it can’t measure belt tension, and it can’t show you a defect that only appears at a Z height above where the print failed. Two things dramatically improve accuracy: a sharp, well-lit, in-focus shot taken close to the defect, and context — material, printer, nozzle size and the failure height. Feed those in and the diagnosis tightens considerably. Our detailed walkthrough on how to diagnose a failed print from a photo covers exactly what to capture, and the defect identification tool guide maps symptoms to fixes end to end.

FAQ

How do I tell warping from delamination in a photo?

Warping lifts the corners and edges off the plate — the base deforms upward. Delamination is a clean horizontal crack partway up the wall, sometimes with a slight upward bend at the split, while the base stays stuck. Warping is an adhesion problem; delamination is a layer-bonding problem, fixed by raising hotend temperature and controlling cooling.

Can a photo diagnose wet filament?

Not directly, but the visual clues are strong: stringing, popping/bubbling texture, rough or inconsistent extrusion and weak layer bonding all point to moisture. Because wet filament is one of the three most common root causes — alongside Z offset and bed temperature — it’s worth ruling out early by drying a sample and reprinting.

Why does the same defect at the same height mean it’s mechanical?

Slicer settings apply across the whole print, so a setting fault produces a defect that scales or repeats everywhere. A fault that recurs at one exact Z height is failing at a physical location — a bind point, a crash into a curled edge, or a lead-screw irregularity — so it’s mechanical, and no profile change will fix it.

Is stringing the same as over-extrusion?

No. Stringing is thin hairs left between features from filament dragged during travel moves, driven by temperature, retraction and moisture. Over-extrusion is too much plastic overall, showing as bulging walls, blobs and dimensional error. They can co-occur, but the fixes differ — tune retraction and temperature for stringing; reduce flow and calibrate E-steps for over-extrusion.