How to Prove Bag Dimension Stability with Process Capability Data

A process capability index is meaningful only after the bag characteristic, tolerance, measurement method, production conditions, sampling order, and statistical model are defined. A high Cpk from a short hand-selected run does not prove stable production. First verify the measurement system and time behavior; then calculate and interpret capability.

Machine Accuracy and Process Capability Are Different Questions

Machine accuracy asks whether a controlled mechanism can reach a commanded position under stated conditions. Process capability asks how the complete production result varies relative to a product specification. PP nonwoven behavior, roll winding, tension, print repeat, tooling, speed, operator setup, temperature, and measurement all contribute to finished bag dimensions.

A Supplier may demonstrate a stable feed mechanism, while the final gusset still varies because the web folds differently across material lots. Keep mechanism tests and finished-product capability studies separate.

Define the Characteristic Without Ambiguity

Bag length sounds simple until one inspector flattens the bag, another includes a fold, and a third pulls the flexible fabric taut. The specification should include datum, measurement path, specimen state, applied flattening or force, instrument, resolution, environmental conditioning if required, and upper and lower limits.

Prove the Measurement Process Before Studying the Line

If two inspectors measure the same flexible bag differently, process indices mix production variation with measurement variation. Check instrument resolution, calibration status, fixture, method, repeatability, operator reproducibility, sample handling, and rounding.

ISO 22514-7 addresses capability of measurement processes and explains why measurement uncertainty must be adequately small before capability or performance indices are meaningful. Use the current licensed edition and qualified statistical support for formal studies.

Design a Study That Includes Time

Twenty consecutive bags after an expert adjustment describe a short moment. They do not include roll-diameter change, warm-up, normal stops, operator intervention, splice passage, or material variation. Sample in production order and retain the time sequence.

• Identify machine, serial number, tooling, software, and recipe.
• Record fabric Supplier, lot, roll, width, GSM, and print condition.
• Include defined low, normal, and high operating states where commercially relevant.
• Sample across the run rather than selecting attractive bags later.
• Mark adjustments, stops, roll changes, alarms, and maintenance on the data plot.
• Keep product families separate unless a justified model combines them.

Look at the Plot Before Cp or Cpk

Plot measurements in production order. Check for trends, cycles, shifts, outliers, mixtures, and changes in spread. A distribution made from two different recipes may look wide even though each recipe is internally consistent. A steady drift with roll diameter points to a different problem than random variation.

Interpret Capability and Performance with the Right Model

Cp commonly compares potential spread with a two-sided tolerance when the process is stable and the model assumptions apply. Cpk also reflects off-center operation. Performance indices such as Pp and Ppk are often calculated using longer-term variation. Names and formulas should follow the organization’s chosen standard and software.

ISO 22514-2:2026 covers capability and performance statistics for time-dependent process models and recognizes different distribution types. Do not force non-normal, mixed, or unstable data into a normal-distribution formula merely because the software offers Cpk.

A Capability Number Does Not Replace Defect Control

Capability describes a measured characteristic under the study conditions. It does not validate seam strength, handle function, print adhesion, contamination, or safety. A centered bag length can coexist with weak welding. Maintain separate control plans for attribute defects and functional tests.

Also preserve the raw data. A buyer should be able to see sampling order, excluded points, reason for exclusion, measurement-system status, statistical model, and software settings. Reporting only Cpk prevents an independent review.

Use Capability in FAT and Improvement Without Gaming It

Before FAT, agree on product, material, tolerance, study duration, sample plan, speed, allowed adjustments, measurement method, and index interpretation. State whether the study measures a short-term machine condition or a broader production process. Do not change limits after seeing the data.

Send Zhengxin your bag drawing, measurement method, ordered data, material and recipe identity, speed history, and adjustment log. We can help map variation to the feed, web, folding, tracking, handle, ultrasonic, or cutting system of the delivered Manufacturer configuration. Formal statistical acceptance should be reviewed by the buyer’s quality specialist.