Short answer
Managing quality in a print farm comes down to seeing failures early and recording the cause of every one. The most common failures are first-layer and adhesion problems, spaghetti, parts coming off the plate, layer shifts and nozzle clogs. AMENT Hive's optional camera watch sends a photo alert for suspicious prints but never stops a print on its own; it reports failure rates by printer, product, filament and hour, and reminds you of maintenance by print hours.
Key takeaways
- Most failures start in the first layer; a short check in the first minutes is the cheapest quality measure there is.
- Give every failed print a reason; without reasons, a quality report is guesswork.
- Hive's camera watch is optional, alerts after two suspicious snapshots and always leaves the decision to a person.
- Read the quality report by printer, product, filament and hour once a week; problems usually cluster on one axis.
- Plain-language error codes and the paused-with-error alert are live but still being validated on real printers.
The most common 3D print failures in a print farm
On one printer a failure is an annoyance. Across twenty printers, the same failure means hours of machine time and a few hundred grams of filament every day. Recognising failures lets you respond correctly and record the cause correctly. The table below sums up the types farms see most and where to look first.
| Failure | What you see | Common causes | First check |
|---|---|---|---|
| First layer / adhesion | Corners lift, lines do not fuse, or nothing sticks at all. | Dirty plate, wrong Z offset, plate too cool. | Wash the plate and watch the first layer; see our bed adhesion guide. |
| Spaghetti | A curly tangle of filament extruded into thin air. | The part came loose or a support failed, so the nozzle prints on nothing. | Find why it came loose: adhesion, support or cooling. |
| Detached part | The part has moved, or sits off the plate entirely. | Weak adhesion, the toolhead striking the part, warping on tall thin parts. | Plate surface, part orientation, a brim. |
| Layer shift | The top of the part is offset sideways, like a step. | Loose belt, a toolhead that catches, too much speed or acceleration. | Belt tension, clean rails, signs of a collision. |
| Clog / under-extrusion | Thin, gappy or missing layers. | Damp filament, a dirty nozzle, heat creep. | Dry the filament, check the nozzle; see nozzles and hotends. |
| Blob on the nozzle | Molten plastic wrapped around the hotend. | Printing on a detached part for a long time. | Stop at once, clean it hot, check the silicone sock. |
For detailed fixes to each symptom, see the troubleshooting guide. Here the focus is on seeing these failures early in a farm and reducing them systematically.
Why record the reason for every failed print?
"We had lots of failures this week" is a feeling. "Adhesion failures on B4 tripled this week" is a finding. The difference is whether each failed print gets a reason. A farm without reasons lives through the same failure on different machines again and again, and nobody connects the dots.
In Hive this is done on the Shop floor screen on the operator's phone. When a plate finishes, the operator taps Done and enters how many units came out good; units that did not are planned again. If the print failed, they tap Failed and pick a reason from a list:
- Did not stick to the plate
- Spaghetti / came loose
- Nozzle clogged
- Layer shift
- Filament ran out
- AMS / feed problem
- Power or network cut
- Operator cancelled
- Wrong file or material
- Other (with an optional note)
The list is short on purpose: an operator standing at the printer with a scraper in one hand should be able to choose in ten seconds. Hive never marks a print done by itself; when the printer reports the end, the card asks the question and a person answers it.
After the reason, the good count is the most valuable piece of information. If a plate holds eight parts and two have surface defects, the print looks successful but the yield is 75 per cent. Hive shows that gap separately in the quality report as good-unit yield, and the two missing units go back into the plan so the order can be completed. The customer does not get a short order, and the report shows the real picture.
How does camera-based print failure detection work?
A print that fails overnight or over lunch can pour filament into the air for hours before anyone notices. A camera watch shortens that time. In Hive, camera watch is optional: it is off by default and only the farm owner can turn it on.
- 1
Snapshot
While a printer with a camera prints, the Farm Connector takes a snapshot about 3 minutes after the start, then another at the set interval (about every 5 minutes by default).
- 2
The AI looks
Each snapshot goes to a vision model. It looks for signs such as spaghetti, a part that came loose, a layer shift, a blob on the nozzle, a clog or an empty bed, and says how sure it is.
- 3
Two-snapshot rule
One suspicious snapshot does not raise an alert. Only when two snapshots in a row of the same print look suspicious above the confidence threshold is an alert opened.
- 4
Photo alert
Everyone in the farm is notified: which printer, which file, what kind of issue and how sure. Open the alert and you see that print's snapshots.
- 5
A person decides
Hive does not pause or stop the printer. Someone goes to the printer, looks, and answers the alert with Really failed or No problem.
The answers tune the watch a little for that printer: No problem answers raise the threshold slightly, confirmed failures lower it slightly. A machine whose camera angle or lighting is misleading produces fewer false alarms over time.
Limits you should know
- It needs a camera: Hive gets camera images from Creality, Bambu A1/P1 and Moonraker printers. It does not currently get images from the Snapmaker U1 or the Bambu X1/H2 series.
- It does not decide: the watch only reads and tells people. That is deliberate: stopping a good print because of a false alarm can cost more than a failure noticed late.
- AI budget: each check counts against the farm's monthly AI budget; when it runs out, checking pauses. Checks per hour can be capped too.
- Short retention: only the newest few snapshots of each print are kept, and they are deleted 7 days after the print ends.
Quality reports: failure rate by printer, product, filament and hour
Every Done and Failed that gets recorded becomes part of a report on Hive's Quality page. The overview shows prints, failure rate, good-unit yield and print hours. The valuable part is the breakdowns: failure rate split by printer, product, material, colour and the hour the print started, with the top failure reasons listed separately.
Each breakdown answers a different question. If one printer's row is dark, the problem is probably that machine: nozzle, belts, plate surface. If one product fails on every printer, look at the file, orientation or supports. If a filament or colour stands out, suspect moisture, a batch difference or the wrong setting set; our storage and drying guide is a good start for the moisture side.
The hour breakdown is often the most surprising. If prints started at night fail more, the cause may be a colder room, or plates prepared in a hurry at the end of the day. The report does not tell you why, but it tells you where to look.
Error codes in plain language and the paused-with-error alert
Printer error codes are often just a string of digits, and a code that lands on your phone at night tells nobody what to do. Hive turns these codes into plain English or Turkish: what the problem is, what to do, and what to do after fixing it.
- Bambu Lab: Bambu's own error text and a link to its wiki, plus a short Hive explanation for common cases (AMS filament run-out or tangle, nozzle clog, plate, first layer, spaghetti, open door).
- Klipper and Creality: the same approach, a plain explanation of the error message the printer reports.
- Paused with an error: when a Bambu printer pauses a print itself because of a problem, such as running out of filament, you get a notification that says what it means and what to do.
When you hit an error you can also ask the assistant: "Why did B6 stop?" It explains the error and shows the relevant button, but it cannot command the printer. Remote pause, resume and stop are only possible with an owner's or production manager's confirmation in the app; the server side of these commands is ready, the buttons are being built for web and iOS, and they have not yet been tested on real printers.
Maintenance reminders by print hours
Some failures are really deferred maintenance: a loose belt turns into a layer shift, a worn nozzle into under-extrusion, a dry bearing into vibration. Calendar-based maintenance does not work well in a farm, because one printer may run a hundred hours a week and another ten. Maintenance should be measured by use.
Hive comes with ready-made maintenance tasks for each printer model: nozzle check, cleaning carbon rods or rails, belt tension, lubrication, PTFE tube check, AMS or CFS cleaning, extruder gears and washing the build plate. Each task counts print hours, number of prints or kilograms of filament. On a printer you might see a reminder such as "B3 nozzle check, 300 hours".
- Status: when a counter reaches 90 per cent of its interval it shows as due soon, past 100 per cent as overdue, and a notification goes out.
- Done: when you finish the task, tap Done; the counter resets and the action is logged.
- Your intervals: change intervals, apply them to every printer of the same model, or add your own tasks.
- Maintenance mode: a printer in maintenance gets nothing planned or started. Only a person turns maintenance mode on and off.
A practical quality routine for print farms
Tools only help when they are part of a routine. The rhythm below works for small and mid-sized farms, and everyone on the team knows their part.
- 1
Every print: watch the first layer
For a new product or a freshly loaded spool, watch the first layer at the printer. Most failures show here, and this is where they are cheapest to fix.
- 2
Every plate: record the result
On the Shop floor screen enter the good count with Done, or the reason with Failed, then tap Plate cleared. Those thirty seconds are the foundation of the whole report.
- 3
Every day: answer the alerts
Do not leave camera alerts open; answer Really failed or No problem. The answers improve both the report and the threshold.
- 4
Every week: read the report
Look at the printer, product and hour with the highest failure rate. If one axis stands out, make a single change and look again a week later.
- 5
Every week: close the maintenance list
Do the tasks marked due soon or overdue and mark them done. A week that starts with overdue maintenance tends to become a week of failures.
The quality module is live in Hive. If you want to set up this routine in your farm, see the features on the Hive page or compare plans and start a 14-day free trial.
Frequently asked questions
What causes spaghetti in 3D printing?
Spaghetti is the nozzle extruding filament into the air with nothing to land on. It usually starts when a part comes loose from the plate, a support fails or a layer shifts. The fix is not the spaghetti itself but the adhesion, support or cooling problem that caused the part to come loose.
Does Hive stop a print when it sees a failure?
No. Hive's camera watch only finds suspicious snapshots and sends a photo alert. A person who looks at the printer decides whether to pause it; Hive never stops a printer on its own.
Which printers does camera watch work with?
Printers Hive can get camera images from: Creality, the Bambu Lab A1 and P1 series, and Klipper printers with a Moonraker webcam. The Snapmaker U1 and the Bambu X1/H2 series do not currently provide camera images to Hive. The Farm Connector also needs to be up to date.
What data is the quality report built from?
The report is built from the plates the team closes on the Shop floor screen (good units or a failure reason) and from prints the printers reported as cancelled or ended with an error. Failure rate is split by printer, product, material, colour and start hour.
How are maintenance intervals set?
Hive comes with ready-made tasks and default intervals for each printer model, counted in print hours, number of prints or kilograms of filament. You can change the intervals from your own experience, apply them to every printer of a model, or add new tasks.