Guard-banding is one technique to protect against in-correct conformity decisions caused by measurement uncertainty or entity dispersion, where the region of permissible values of the entity’s quality characteristic is reduced in proportion to the actual measurement uncertainty or dispersion.
A ‘guard-band’ is placed around each specification limit with a width equal to some fraction of the actual measurement uncertainty.
According to the guard-band rule, to protect against consumer risk, conformity to specification will be indicated if the result, ym, of a measurement of the quality characteristic of the entity under test is in the region of acceptable values bounded by:
where the dispersion in quality characteristic values is σ and equals the (standard) measurement uncertainty, u, in specific conformity assessment of one item of an entity. The ‘guard-band’ factor, h, defines which fraction of the dispersion is used for the width of the band.
Correspondingly, to protect against producer risk, non-conformity to specification will be indicated if the result, ym, of a measurement of the quality characteristic of the entity under test is in the region of acceptable values bounded by:
Guard-banding is illustrated in the figure:
Guard-banding [© ISO Figure 5 of ISO 14253-1:1998]
- ISO 14253-1:1998, Geometrical Product Specification (GPS) — Inspection by measurement of workpieces and measuring instruments — Part 1: Decision rules for proving conformance or non-conformance with specifications.
Consequences of guard-banding
Figure: Economics of guard-banding: example prepackaged goods [Pendrill 2009]
As with statistical ‘significance’ testing and rules of ‘thumb’ about capability factors, the choice of size of guard-band (factor h) in purely % risk terms can appear largely arbitrary. Adding measures of impact and cost to the choice of guard-band goes some way to removing this arbitrariness – see above figure.
L. R. Pendrill 2009, ”An Optimised Uncertainty Approach to Guardbanding in Global Conformity Assessment”, Advanced Mathematical and Computational Tools in Metrology VIII in Data Modeling for Metrology and Testing in Measurement Science Series: Modeling and Simulation in Science, Engineering and Technology, Birkhausewr (Boston), 2009, ISBN: 978-0-8176-4592-2 http://www.worldscibooks.com/mathematics/7212.html