Control Charts for Attributes - Statistical Process Control
Attribute data assume only two values, such as pass/fail or good/bad: they are counted rather than measured. Many quality characteristics can only be monitored in this way, such as the soundness of a soldered connection or the absence of components on a printed circuit board. Before proceeding further, a distinction must be made between defectives and defects. The term defective refers to a non-conforming item of output. A defect is a non-conforming feature of a single item of output: such an item may have more than one defect. Control charts for attributes have several advantages:
- Collection of attribute data does not require measurement equipment and the skills to use them
- They can be used to monitor any process, since the process output can be judged as either conforming or non conforming
- Different attributes may be recorded on a single chart.
On the other hand, they have the following disadvantages:
- Sample sizes are generally larger than for variables charts. Ideally, the sample size should be chosen such that, on average, at least one defective/defect is found in the sample
- Attribute data do not indicate degrees of defectiveness: a non conforming item may be slightly defective or greatly defective.
Several attributes control charts have been devised, but those most commonly used are those described below. Two charts that look for defective items are discussed - the p-chart that looks at the proportion of defectives when the sample size is variable, and the np- chart that looks for the proportion of defectives in a constant sample size. Another sort of chart - the c-chart - is also described. This looks for the number of defects on an item.