ProductionNewVerified Real deployment at a named or precisely described company, with a measured result from a primary source.

AI reads each molding cycle and tells the setter which machine setting to adjust, so scrap and start-up losses fall (Beko, Zahoransky)

Added 2026-10-21 · Last updated 2026-10-21

Role

Production (injection molding: process setting and shift operation) · Typical users: process technician or setter, molding supervisor, production manager, process engineer, quality engineer

The problem

An injection molding recipe that was right at sampling drifts in series production: recycled material varies from batch to batch, humidity and temperature change, and the same machine, mold and material do not give the same part every day. Short shots, flash, sink marks and weld lines appear, and the fix depends on how fast and how well the setter on shift reacts. Experienced setters are hard to find and slow to train. At a global medtech manufacturer quoted in the Zahoransky article, changing material batches caused significant scrap and up to 30% swings in OEE between shifts and batches; the Beko researchers note that scrap rates vary with operators' competence and their ability to intervene quickly.

How AI helps

Sensors capture every cycle: cavity pressure curves at Beko, and at Zahoransky the machine and peripheral data (hot runner, dryer, temperature control units), the properties of the material batch in use, ambient conditions and quality test results. A model learns which pattern belongs to good parts and, when a cycle drifts out of that window, recommends which setting to change and in which direction (injection pressure, speed or time, holding pressure or time, shot volume at Beko; up to 60 setting values at Zahoransky, always inside the validated limits). The setter or process engineer confirms the change; at Zahoransky the vendor says the tool can also write settings back to the machine within the validated windows.

Step by step

1. Pick one or two parts with high scrap, long cycles or painful start-ups, as Beko did, and record the baseline: scrap by defect type, start-up scrap, cycle time and downtime caused at assembly.
2. Install or connect the data: cavity pressure where possible, machine and peripheral data, material batch and ambient conditions, and the quality result of each cycle or lot.
3. Define the good window with process engineering: run a design of experiments at the limits and keep only settings that give good parts, including hidden defects such as porosity.
4. Run in advice mode: the system suggests the change, the setter decides and the result is logged; measure how often the advice was right.
5. Keep the validated limits locked; in regulated production, document the model and the change process before allowing any automatic write-back.
6. Track scrap, start-up scrap, cycle time and OEE swings between shifts each month, then extend to similar parts.

Tools

• Cavity pressure sensors (Kistler) with a custom neural network built by Beko and Koç University: used at Beko
• plus10 Hopper (live-learning setting recommendations) and Shannon (operator assistance during stops): used at Zahoransky
• Other tools to evaluate: your machine maker's assistance systems (for example Arburg, Engel, KraussMaffei), cavity pressure monitoring (Kistler ComoNeo, RJG eDART), and AI process tools such as Gauss ML Optimyzer
• Any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below, on a shift's cycle data and defect log, to prepare a troubleshooting sheet for the setter

As reported by the source; most figures are vendor or self-reported and not independently verified.

Human review: cautions

• The AI advises within the process window; the setter or process engineer makes the change, and any automatic write-back must stay inside validated limits.
• Defining the good window takes engineering work (simulation and designed experiments at Beko) before the model can help.
• In medical and other regulated production, a model that changes settings is part of the validated process: document it and its change control.
• Beko's figures come from a peer-reviewed study co-written by Beko staff; the Zahoransky figures come from customer feedback reported in an article co-written by the vendor.

You are an experienced injection molding process technician.

I will paste: (1) the validated process window for this part [parameter, set value, lower and upper limit]; (2) cycle data from the shift [time, cycle number, cycle time, cushion, injection pressure peak, holding pressure, mold and melt temperature, and cavity pressure peak if available]; (3) the defect log [time, defect type (short shot, flash, sink mark, weld line, burn, warpage, dimension out), quantity]; (4) the material batch in use and its drying record, and ambient temperature and humidity if recorded.

Do this:
1. Line up the defects with the cycle data and list the readings that changed before each defect appeared, quoting the values.
2. For each defect type, list the usual causes and which ones the data supports or rules out.
3. Propose at most three setting changes, each with direction (increase / decrease), the reason, and confirmation that it stays inside the validated window; never propose a value outside it.
4. Say what to check first that is not a setting (material moisture, batch change, mold temperature unit, worn check ring, vent blockage).
5. List the data that is missing to be sure, as questions.

The setter decides and makes any change on the machine.
Use only the data I paste in; if something is missing, write 'not found' or list it as a question instead of guessing. Do not send, order, post or change anything in any system: produce drafts for my review and approval.

Related cases

• Anomaly detection on machine and process data to cut scrap
• Machine learning recommends the minimum alloy addition for each heat at a steel mill (Gerdau, Fero Labs)
• Operator assistant over SOPs, troubleshooting guides and process data
• AI rewrites the production schedule in minutes when a line slows or an order changes

All 9 Production use cases →

Sources

• https://link.springer.com/article/10.1007/s00170-025-15611-x
• https://en.kunststoffe.de/a/specialistarticle/smart-machine-ramp-ups-stable-series-pro-7129063