About this lesson
Control Charting Process
There is an eight-step approach used to establish control charts for a process. Part of that process is the selection of a control chart type based upon the nature of the data being monitored.
When to use
When creating a control chart for the first time on a process, use this eight-step process.
Control charting can be implemented on a process using this eight-step process.
- Select process characteristic to control. This should be a characteristic that was identified as critical during the design process or one that is a clear signal for conforming process performance.
- Determine appropriate control chart. This is determined by the characteristics of the data. The chart below is a decision tree for selecting control charts.
- Collect data and calculate appropriate statistics. Create or use a a data stream from the process. Once you have 30 data points, calculate descriptive statistics for the data set including the mean and standard deviation. Using these and the appropriate formulas and constants, calculate the control limits.
- Construct preliminary control charts. This will include the data, the mean, and the control limits that have been calculated.
- Establish control. If there is the presence of special causes, find and eliminate them. Once the only variation shown is common cause variation, you are ready to move to the next step.
- Determine the process capability (Cpk). If necessary, center the process to make the best use of the available span between the spec limits. If the Cp index shows that you have a poor process sigma, then consider changing the process and reduce common causes variation.
- Once the process is capable and stable, construct the final control charts. Be sure that you only include data from the point where the special cause variation has been eliminated and after any process changes to reduce common cause variation have been implemented.
- Use the control charts for ongoing process control purposes.
Hints & tips
- When selecting a control chart, be clear on whether the data is variable or attribute and if it is attribute you know whether it is defects or defectives that are being counted.
- If the process is a low volume process, use all data points or the smallest sample size consistent with the Central Limit Theorem. High volume processes should be sampled at a rate that is practical.
- If a process is struggling or having performance problems, sample more frequently until the sources of instability are resolved.
Lesson notes are only available for subscribers.