Control Chart: How to do it
The Quality Toolbook >
Control Chart > How to do it
When to use it | How to understand it |
Example | How to do it | Practical
variations
<-- Previous |
Next -->
How to do it
- Identify the objective of using the Control Chart. Typically this will be
either to detect defects or to monitor a suspect or critical process. For
example, 'to monitor the accuracy of a breakfast cereal packet filling process'.
The process selected should repeat sufficiently often to provide enough
measures to be able to plot a Control Chart.
- Identify the actual measurement to be made, including what to measure, and
where in the process to measure it. Select the measurements based on their
ability to identify problems or defects.
Focus the measurement to minimize the likely variation. For example, a
separate Control Chart for each of several identical canning lines will be more
likely to identify problems on one line than using one chart for all the output
of all lines, where significant variation on one line may be swamped by measures
from the other lines.
Reduce waste by identifying the measurement at the earliest point in the
process where problems may be detected.
The actual measurement should be reasonably easy to carry out.
- Identify the type of Control Charts to use. This will depend on the type of
measurement being made. For more details on selecting the right Control Chart
for your application, see Choosing the type of Control
Chart.
- Choose the subgroup. This is the group of measurements
that will make up each plotted point on the Control Chart.
Each subgroup typically contains the same number of measurements, although p-charts
and c-charts are bounded by events, such as time (e.g. measurements per week),
people or batches. In any case, there should be enough measurements in each
subgroup to make the chart statistically correct and Table gives guidelines for
this.
The subgroup should be selected with the aim of (a) making the measurement
within each subgroup as consistent as possible, whilst (b) maximizing the
chance of highlighting differences between subgroups. Considerations
include:
- Synchronizing measurement points with other process variables. For example,
measure weekly rather than every four days.
- Using experience to determine subgroups, for example, known tool wear rates.
- Using larger subgroups, as they result in Control Charts which are more
sensitive to change.
- Using smaller subgroups when they are expensive or time-consuming.
- Measuring more frequently when significant variation can occur over a short
period.
- Initially measuring more, then reducing measurements as the data is
understood.
- Using consecutive measurements, rather than a random sample, as this will
result in less variation within the subgroup, with tighter, more sensitive
control limits.
- Selecting subgroup measurement which seldom results in zero value points.
For example, counting customer complaints per hour when there are only one or
two per day, will give many points plotted on the zero line.
- Prepare for measurement. This should aim to make measurement as simple and
error-free as possible. If possible, automate the measurement process. If
measurements are to be collected by hand, design a data collection form that
eases both the collection and the subsequent calculations.
Ensure that the people involved with the collection of data and construction
and interpretation of the Control Chart are able to perform their tasks
efficiently and accurately. Train them as appropriate, including the use of
practical trials.
- Make the measurements as planned in step 5.
- Calculate mean and upper and lower control limits, using the links below for
the appropriate Control Chart.
Note that although most control limits are straight lines, the p- and u-
charts may have control limits that are different for each plotted point.
Sometimes a lower control limit is calculated as less than zero, but as it is
impossible for the plot to go below zero, the lower control limit is set at the
zero line.
- Draw the chart. This should include:
- One plotted point for each subgroup, with a line drawn between successive
points.
- Horizontal lines for each of the central line, upper control limit and
lower control limit.
- Labeling and other information to uniquely identify the chart and help
with any subsequent investigation.
- Interpret the chart, looking for significant patterns and points, and act on the results. Typically this will involve finding for the cause of
any identified significant set of points, followed by devising a method of
correcting the problem. This can require a fair amount of effort, as the Control
Chart shows when to look, but not where.
<-- Previous |
Next -->
|