Process Capability Indices, Alternatives to

Abstract

Abstract Since we (the statistical community) are to blame for the introduction of process capability indices (PCIs) and their ensuing use and abuse in the industrial world, it is now up to us to provide an alternative to PCIs that is both statistically sound and relatively simple to understand. The problems with the present capability indices have been well documented elsewhere (e.g., Gunter 1, Kotz 2, Journal of Quality Technology (JQT) 24 3) and these will not be addressed here. Instead, the dual roles of specifications and PCIs in process and product development and in the manufacturing environment will be discussed. A basic approach to meet most of the goals of PCIs without some of their problems will then be introduced. Incorporated into the measures will be the concepts of confidence level and sample size. An example of this procedure is included. The relatively simple assumptions of this approach can then be modified to meet requirements where PCIs are presently being applied. These measures are intended to be a foundation rather than a total solution to all the needs of PCI users. However, these techniques will provide a direction for future work.