A paper cup process change point is a time at which the level or variation of a measured response appears to shift. Searching many possible split points can find an apparent change even in random data, so statistical evidence must be combined with machine, material and measurement records.
This article supports structured investigation, not automatic statistical proof or machine adjustment. A competent analyst and process engineer must choose the method and interpret operational evidence. Preserve the original sequence, excluded observations and analysis choices so another reviewer can reproduce the conclusion.
Diagnostic decision table
| Pattern | Evidence to add | Risk |
|---|---|---|
| Sudden step | Alarm, adjustment or material event | Assuming the nearest event caused it |
| Gradual drift | Temperature, wear and conditioning trends | Forcing one exact split point |
| Single spike | Sample and measurement validity | Splitting the whole run around an outlier |
| Variance change | Raw data and subgroup structure | Reviewing only averages |
Preserve original production order
Retain timestamp, forming position, material lot, operator, machine state and sample identity. Do not sort results by value before plotting. Identify gaps, delayed entries and samples whose timestamp reflects testing rather than production.
Use a response with a stable definition and measurement method. A method or fixture change can create an apparent process change. Keep raw readings and flag invalid observations through predefined rules instead of deleting inconvenient points.
Plot response and process events together
Display individual results or rational subgroups in production order. Annotate starts, stops, splices, maintenance, setpoint changes, alarms and environmental events. Review both center and spread.
Avoid drawing a vertical line at the most visually attractive location and declaring cause. The operational event may precede the measured response because of transport delay, mixed material or sampling frequency. Estimate that lag from the process.
Use a preplanned or defensible analysis
If a known intervention time exists, define the before-and-after comparison before viewing outcomes. For an unknown change, use a suitable change-point method with control for repeated searching and confirm assumptions with a statistician.
Distinguish step changes, trends, cycles and isolated outliers. Do not run many algorithms and report only the smallest probability. Include uncertainty in the estimated location and test sensitivity to missing or autocorrelated data.
Investigate competing explanations
Check measurement checks, calibration, operator, material, forming position and downstream handling. Compare related responses that should move if the proposed mechanism is real. A heating change may affect sealing evidence differently from a packaging event.
Use retained cups or controlled follow-up tests where available. Temporal proximity is evidence for investigation, not proof of causation. Avoid reversing a necessary safety or quality action merely because the chart shift has not yet been fully explained.
Confirm with prospective monitoring
After an authorized correction, define expected behavior and collect new time-ordered data prospectively. Confirm that the change persists across representative production and does not create another defect. Keep the historical boundary visible.
Send HANNAI the time plot, event log, measurement checks and proposed mechanism. This supports focused machine review while avoiding a retrospective split that exaggerates certainty.
Related equipment and next checks
Review the HN-S120 paper cup making machine and the related guide to control chart subgroup guide. The Engineering Notes archive connects these checks with wider machine planning.
Frequently asked questions
Does a visible step prove a process change?
Confirm it with measurement, statistical and operational evidence.
Can data be sorted before analysis?
Preserve production time order for change-point investigation.
Is the closest machine event automatically the cause?
No. Consider lag and competing material or measurement explanations.
How is a correction confirmed?
Monitor new representative production using a prospective plan.
Review the evidence with HANNAI
Send your machine model, component identification and the observations described in this guide. Include the drawing or material reference and any before-and-after measurements so the required next check can be identified.