Implementing SPC in a plant
A practical rollout plan: choosing characteristics, measurement, training, reaction plans and sustaining the system.
1. Choose what to chart
Start with a few characteristics that matter: special characteristics, those with scrap or complaint history, and process parameters that drive them. Charting everything overwhelms people and gets ignored.
2. Fix the measurement first
Run Gage R&R on each measurement. Check resolution. A chart of a poor measurement is a chart of noise.
3. Design the sampling
Decide subgroup size, frequency and who measures. Use rational subgroups. Write it in the control plan.
4. Build a baseline
Collect 20 to 25 subgroups, set limits, investigate signals, and lock the limits.
5. Write the reaction plan
For each chart: what to do on a signal, who to call, when to stop the line, how to contain parts, and where to record the cause. Without a reaction plan, a chart is decoration.
6. Train
Operators: how to take samples, plot, recognize signals and follow the reaction plan. Engineers and supervisors: subgrouping, interpretation, capability and investigation.
7. Review and improve
Review charts and cause logs regularly. Use capability results and Pareto of causes to choose improvement projects. Retire charts that no longer add value and add new ones as processes change.
Common reasons SPC fails
- Charts filled in after the fact, or limits copied from tolerances.
- No reaction plan, or signals ignored.
- Poor measurement systems.
- Too many charts and nobody looks at them.
- Management treats a signal as blame instead of information.
Questions and answers
Should operators or engineers own the charts?
Operators should take the data and react first; engineers should review trends, investigate recurring causes and improve the process. Both need to see the charts.
