Statistical Process Control Based on Attribute Data Obtained by a Narrowed Tolerance

Abstract

Process control by means of a p-chart or np-chart becomes almost impossible, if the quality level of the process in question is very good. However, if the quality characteristic is measurable then by narrowing the given tolerance limits and using accordingly changed gauges one can easily control the process by means of an attribute control chart. In this paper a method is developed to determine an optimal (in a certain sense) way to select the new narrowed tolerance tolerance limits if the distribution of the quality characteristic can be sufficiently well approximated by a normal distribution.

This paper is the outcome of engineering research aiming at obtaining a useful result which can be applied at shop-floor level. A majority of modern methods (including ISO 9000 ideology) try to prevent defects by searching for defects. In contrast, the idea here is to increase sensibility of the method by defining a “high virtual defective level”, which is used for defect prevention without actually searching for defects.