A precision moulder only reacted when a control chart crossed a limit, by which time parts were sometimes out of tolerance. Trend analysis showed a critical diameter grew predictably as an insert wore; a trend-based action limit cut out-of-tolerance events from nine a year to one.
Key numbers
A precision moulder had done everything the textbook asked. Its critical diameter had an SPC chart at the press, operators recorded measurements faithfully and reacted when a point crossed a control limit. Yet about nine times a year, parts were found out of tolerance. Each event meant sorting, scrap, a tool pulled from the press without warning and an awkward call to the customer.
The situation
The control chart was doing its job: detecting change. The problem was timing. By the time the chart signalled, the process had already drifted far enough that some parts were outside the specification. Unplanned tool pulls, 14 a year, disrupted the schedule and the toolroom.
What the data showed
The team pulled years of historical chart data and plotted the diameter against the shot count of the core insert, rather than against time.
A clear pattern emerged. The diameter grew slowly and predictably as the core insert wore, accelerating towards the end of its life. The relationship was consistent across insert life cycles. The control chart could not see it, because it looked at each point against limits rather than at the trend across hundreds of thousands of shots, the limitation discussed in SPC without the jargon.
What changed
- A trend-based action limit was set at 12.030 mm, inside the specification limit, giving enough warning to plan an insert change before parts went out of tolerance.
- Insert replacement was planned when the trend approached the action limit, scheduled into production rather than forced by a failure.
- Spare inserts were held in stock for the three most critical tools, so replacements could be made quickly.
- Shot counts were recorded with the SPC data, so the trend could be tracked on every tool.
Results
Out-of-tolerance events fell from nine a year to one. Unplanned tool pulls dropped from 14 to three, a reduction of almost 80%. Because inserts were now replaced on evidence rather than in panic or on a conservative fixed schedule, the share of insert life used rose from 72% to 91%.
Why it worked
The data had been there for years. The difference was looking at it against the right variable, shot count rather than time, and acting on the trend before it reached the limit. Prediction replaced reaction.
Lessons for your operation
- Store SPC data with context: tool, cavity and shot count.
- Plot critical dimensions against shot count to reveal wear trends.
- Set action limits inside specification limits, based on trend and lead time.
- Hold spares for critical wear components so planned changes are quick.
- Predictive maintenance can extend component life, not just prevent failures.
