Page 176 - Six Sigma for electronics design and manufacturing

The Use of Six Sigma with High- and Low-Volume Products and Processes
95% Confidence
= 0.05, therefore the 95% confidence limits are 0.025 and 0.975 @
= 8:
8 · (0.25) /17.535 < < 8(0.25) /2.180
2
2
2
2
or
0.0285 < < 0.229
0.17 < < 0.48
99% Confidence
= 0.01, therefore the 99% confidence limits are 0.005 and 0.995 @
= 8:
2
2
2
8 · (0.25) /21.955 < < 8(0.25) /1.344 145
2
0.0228 < < 0.372 or 0.15 < < 0.61
Note that the confidence interval gets larger as the confidence limits
increase.
5.2 Determining Process Capability
Process capability is the analysis of a process to determine its quality.
A single or several quality characteristics are selected, some of which
might be variable or attribute. For variable characteristics, the distri-
bution of the data collected is for normality, and the distribution aver-
age and standard deviation are calculated. It has been shown in
this and previous chapters that it takes a sample size of 30 measure-
ments to directly obtain these two parameters and determine whether
the distribution of data is normal. For low-volume production, the pre-
vious section discussed methods of determining a confidence interval
for the two parameters. The confidence limits from these intervals
could be used for worst-case determination of six sigma quality. For at-
tribute processes, the defect rate is determined for parts that are man-
ufactured in small quantities as prototypes, or from similar parts in
current production. The reject rate can be translated into DPU (PPM),
DPMO, FTY, Cpk, or sigma quality, as was shown in Chapters 2 and 4.
The amount of sampling required for determining process capabili-
ty is also dependent on whether the process has been in production
(existing) for some time or is a new process is being created. It is also
desirable that once the process is operating on a regular basis, and a
reasonable level of quality is achieved, the quality characteristic(s)
being measured be charted for statistical control in control charts. For
quality level approaching six sigma and beyond, control charting
might not be required; a total quality management program to moni-
tor individual defects per period as opposed to use the sampling meth-
ods of control charts (refer to the discussion in Chapter 3 regarding
this issue) could be substituted.