A US medical device supplier found out-of-spec tubing only at final inspection. Control charts, clear reaction plans and tighter die temperature control raised Cpk from 0.92 to 1.71 and eliminated out-of-spec batches.
Key numbers
A US medical device supplier extruded small-bore tubing with an outside diameter tolerance of ±0.05 mm. Most batches were fine. Some were not, and the bad ones were discovered only at final inspection, after hours of production. Each one meant scrapped material, delayed shipments and an anxious conversation with the customer, who required a Cpk of at least 1.33.
The situation
The process had been run the same way for years. Operators checked diameter periodically and adjusted puller speed by feel when they thought it was drifting. Capability was 0.92, well below the customer's requirement, and seven batches had been rejected in the previous year.
What the data showed
The team first confirmed that the diameter measurement could be trusted, using the measurement system checks described in our introduction to Gauge R&R. Then they analysed historical data run by run.
The pattern was clear: diameter drifted slowly during each run as die temperature rose. Operators' manual adjustments sometimes corrected it and sometimes over-corrected, adding their own variation. The process was not out of control at random. It had a predictable special cause that nobody was tracking.
What changed
- X-bar and R control charts on diameter, with samples every 15 minutes, kept at the line and updated by operators.
- Clear reaction plans for each out-of-control rule: what to check, what to adjust and when to stop and call engineering.
- The die temperature control loop tightened, removing the main source of drift.
- Operators trained to read the charts and, just as importantly, to leave the process alone when it was in control.
Results
Cpk rose from 0.92 to 1.71 over six months, comfortably above the customer's 1.33 requirement. No out-of-spec batches have been produced since. With demonstrated capability, inspection was reduced from 100% checking to sampling, at 20% of the previous frequency.
Why it worked
The control charts did two jobs. They detected the drift early, before parts went out of specification. And they stopped well-meaning adjustments when the process was simply showing normal variation. Fixing the die temperature control removed the main special cause, so the charts had less to react to.
Lessons for your operation
- Prove the measurement system before charting. Otherwise you are charting noise.
- Look for patterns over time within each run. Drift with temperature or wear is common and predictable.
- Reaction plans turn charts into action. Without them, signals are ignored.
- Teaching operators when not to adjust is as valuable as teaching them when to.
- Demonstrated capability can reduce inspection cost as well as scrap.
