When the Settings Live in an Operator's Notebook
Many plants run the same machine with the same material and still see different defect rates between the day and night shifts. Trace it back and the cause is often not the equipment or the material but the settings. Mold temperature, holding pressure, and cooling time are written in an operator's notebook or on a note taped to the machine, passed along verbally at shift change, and typed into the HMI by hand.
That arrangement creates two problems.
Turning Recipes into Master Data
The first step is to move settings out of individual know-how and into company master data.
Recipe Structure and Tolerances
Define recipes per product × machine × process step, because the best conditions for the same part can differ between machine 1 and machine 2. Each parameter gets a target plus upper and lower limits: mold temperature at 60°C ±3°C, for example, or cooling time at 12 seconds ±1 second. A recipe without limits gives you no basis for judging a deviation, so it cannot serve as a control tool.
Versions, Approval, and Change History
Between MES and the Machine: Download and Actuals
Automatic Recipe Download
When a work order starts, the MES finds the approved version for that product and machine and sends it to the equipment over OPC UA or a PLC connection. The operator confirms the values on screen instead of typing them. Removing manual entry structurally reduces typos and the classic mistake of running a new part on the previous part's settings.
Collecting Actual Values and Handling Deviations
What you download and what the machine actually runs can differ, so collect both setpoints and measured values and link them to the lot. When a measured value drifts out of range, respond in tiers.
Interlocking every parameter will hurt availability, so reserve it for the few that directly determine quality.
Golden Batch: Making the Best Lots the Standard
Once data accumulates, you can find the conditions behind your best lots. Suppose you take the last 200 lots and compare the parameter distribution of the top 10% by yield, 20 lots, against the rest. You may find an operating window that is narrower and better than the current standard.
Do not promote that finding straight to the standard. With a small number of lots, the result may be a coincidence driven by a material batch or the season. Register the candidate as a trial recipe version, run enough lots under it, and compare it with the current conditions using control charts and process capability indices. Approve it only if the improvement is significant. The rule of never changing the standard without statistical evidence is what keeps the recipe trustworthy.
Rollout Order and Common Failures
Trying to cover every process and every parameter from day one means months spent on master data alone. Start narrow, with one or two critical processes and five to ten key parameters each, where defect costs are highest, and expand after the results are clear.
The most common failure is locking out shop-floor adjustments completely. Small corrections for material moisture or ambient temperature are a fact of production. Block them and operators switch the machine to local mode, and the records disappear again. A more realistic design allows adjustments inside the tolerance and logs them automatically with a reason, while changes beyond the limits go through approval.
How POLYGLOTSOFT Helps
POLYGLOTSOFT builds recipe management as an MES extension: recipe master data, versioning and approval workflows, OPC UA and PLC integration, deviation detection, and lot traceability. If you already run an MES, we can add the recipe module alone, and our subscription development model lets you begin with your critical processes and expand in stages without a large upfront investment. If some of your settings still live in a notebook, contact POLYGLOTSOFT.