How we change what others think, feel, believe and do
Control Chart (part 2: Types of Control Chart)
Last time we discussed the basic interpretation of Control Charts, including the identification of special causes of variation, shifts, trends and cycles. We now get more technical as we look at the different types of Control Chart that you can use, and when each type should be used.
There are two types of measurement which you can measure and plot on a Control Chart.
· Variables answer the question ‘how much?’ and are measured in quantitative units, for example weight, voltage or time.
· Attributes answer the question ‘how many?’ and are measured as a count, for example the number of defects in a batch of products.
A trick that is used to make Control Charts more sensitive is to combine multiple measurements into a single point (this is called a ‘subgroup’). This makes the the chart more sensitive to variations, but does mean that you need more measurements. A Control Chart should have at least 25 points on it, which can mean several hundred measurements are needed.
Subgroups also mean that you can end up with two charts: one which considers the difference between subgroups and one which considers the difference within subgroups. Thus, for example, if you combined a day’s readings in one subgroup (ie. one point on the chart), then using a single chart might miss variation within a single day (for example warming-up and lunchtime effects). The second chart copes with this problem.
When you are measuring variables, there are three types of Control Chart that you can use (X/MR, X-bar/R and X-bar/S). This decision is based on the number of measurements that you make and consequently how many measurements you can combine into a single point (subgroup). Variables charts are useful for such as measuring machine tool wear and predicting when the tool needs changing before it creates defective products. Variables charts are more sensitive to change than Attributes charts, but can be more difficult both in the identification of what to measure and also in the actual measurement.
A different attribute Control Chart is needed depending both on whether you are counting the number of defects per item or whether you are just counting total defects. Thus, for example, a production line might output 100 televisions, with 100 defects. This could mean that they are all defective or only one television is defective (or anything in between). The right attribute chart is also selected based on the whether there is constant number of measurements in each point (subgroup) on the chart. Thus there are four types of attribute chart to choose from (u, c, p and np).
Attribute charts are useful for both machine- and people-based processes. Data for them is often readily available and they are easily understood. It can thus be easier to start with these, then move on to Variables charts for more detailed analysis.
Fig. 1 shows a decision tree that you can use to identify the type of Control Chart you need.
Fig. 1. Choosing the right type of Control Chart
Next time: Control Chart (part 3: producing the chart)
This article first appeared in Quality World, the journal of the Chartered Quality Institute
And the big