
PROCESS CAPABILITY
Once you have brought a process under statistical control (no
assignable/special causes or other nonnormal distributions of subgroup average plot
points; i.e., there are only common causes of variation remaining), you can calculate a
process capability index. In addition to the process being under statistical control (so
that it is repeatable, or predictable), the raw data must be normally distributed. This is
because you will be using an estimate of the process standard deviation to calculate the
process capability. Standard deviation does not apply to nonnormal distributions (more
advanced techniques provide data transformation to help in these situations). It is not
necessary, however, for the raw data to be normally distributed in order to use SPC
charts. This is due to the Central Limit Theorem.
A Cp
Index is a ratio of the process specification range (Upper Spec Limit  Lower Spec Limit)
divided by 6 standard deviations. A Cpk index will be the same as a Cp if a process mean
is centered on the spec target, and lower if the mean is not centered on target. The
Cp index uses a standard deviation estimated from Rbar. The Ppk index uses a standard
deviation that is calculated directly from the data (and usually
longerterm). The Cpm and Ppm indices use the mean squares
deviation from the target rather than the process average. 