Today
An injection molding process is set up and released at sampling, then it wanders shift by shift. Setters adjust pressures and temperatures, a new material batch flows differently, the dryer weakens, the mold wears. The machine records every shot, yet nobody compares the values with the released window. The first visible sign is often a customer complaint about short shots or flash.
An example
At Polivo Kunststoffverpackungen, the preforms for the BH 500 ml rPET container run on an injection molding machine with a 48-cavity hot runner mold. Since Sunday evening the peak injection pressure has been falling slowly; all values are still within the released window. The agent puts the trend next to the dryer data: over the same period the dew point has risen from −45 °C to −31 °C. It flags the drift with a forecast that the pressure will reach the lower limit in about 14 hours and links a case from April in which moist pellets had lowered the viscosity of the preforms. Maintenance replaces the dryer's regeneration heater, and quality measures the viscosity on the retained samples from the night.
Your data, wherever it lives
Your process data lives today shot by shot in the machine controllers and the MES, the released window in the setting sheet from sampling, material batches in the ERP, dryer values in the dryer itself and test results in the CAQ. demi reads, understands and links them per machine and mold, with your rights. The AI sees drift while all values are still within the window, not at the customer complaint, and setters and process engineering decide what to change.
Where the data lives today
- Machines & sensors
- MES
- Office files
- ERP
- QM system
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
Data sources in this case: per-shot process data from the machine controllers, orders and changeovers from the MES, material batches from the ERP, dryer values, mold history and test results from the CAQ. demi replaces nothing and puts them on one timeline per machine and mold, together with the released process window from the setting sheet.
- 2 · PermitPermission Layer
Setters and shift leaders see their machines, process engineering and quality see all molds and windows. Analysis runs by machine, mold and batch, never by person. Only process engineering may change a process window, and every change is recorded.
- 3 · ActIntelligence Layer
The agent compares every shot with the window and follows cushion, injection pressure, switchover point, melt temperature, mold temperature and cycle time over time. When a value heads for a limit, it alerts with a forecast of when the limit will be reached and shows what changed at the same time: material batch, dryer dew point, regrind share or shots since the last maintenance. The values come from the machine data; the language model only explains the connection.
How it runs
- 1
demi reads the process data of every shot and the mold's released window from the setting sheet.
- 2
The agent checks every shot against the window and follows cushion, pressures, temperatures and cycle time over time.
- 3
When a value heads for the edge of the window, it alerts the shift with a forecast of when the edge will be reached.
- 4
With the alert it shows material batch, dryer values, regrind share and mold condition from the same period.
- 5
Setters and process engineering correct within the window or plan maintenance; quality decides on suspect parts.
What changes
Drift caught before parts leave the window
Every correction documented against the released window
Causes from material, drying and mold at a glance
What does it bring you?
Check it with your own numbers
Every starting value is an assumption you can change. The calculation sits right below, step by step.
Your numbers
Result
+130 hfreed per year
Assumptions and calculation
demi lines up shot data, material, drying and mold condition; you decide on the correction.
3 h → 2.1 h per case
How we calculate
12 root-cause analyses × 3 h × 30% × 1.0 (manual) × 12 months = 130 h
Value per year, before demi costs
Freed per year+130 h
Assumption: labor cost of $48.62 per hour in US manufacturing in June 2026 (BLS), plus workplace, IT and overhead.
= Value per year, before demi costs
+$8,450
gained per year
130 h ÷ 1,760 h per role = 0.1 full-time roles
An estimate, not a promise. The starting values are assumptions for an industrial company with 150 to 500 people.
Combine with more cases and get the packageQuestions about this use case
Where does demi get the released process window from?
From the setting sheet and the process release from sampling, per mold and machine, whether it lives in the MES, in a spreadsheet or as a scan. Process engineering confirms the window once; after that the agent checks every shot against it. If a mold has no window, it says so instead of inventing one.
Which injection molding machines can demi read?
Any that output their process data: through EUROMAP interfaces, through the MES or as a file from the machine computer. For older machines, the export of the cycle data is often enough. How the data gets to demi, we clarify in conversation.
Why do drying and material matter here?
Because many drifts start there. Moist PA, PBT, PC or PET pellets flow differently, a new batch has a different melt flow, more regrind changes the dosing. The agent puts dew point, residence time in the dryer, material batch and regrind share on the same timeline as the shots.
Who adjusts the machine when demi reports drift?
The setter at the machine, within the released window. demi only reads and reports; it changes nothing. Only process engineering releases a new window, and both are recorded.
