First Edition Dated September, 2000 28
As part of Section 4.2.1, the process step, control method used, subgroup
size, and sampling frequency must be recorded on the AQS Control Plan, or
equivalent, or manufacturing plan.
Note: Boeing reserves the right to require that statistical process control
charts be used to control the variation of Boeing identified key
characteristics,. Boeing engineering documentation typically takes
precedence over other documents controlling part or process quality.
However, Boeing may relieve the supplier of maintaining SPC charts on
Boeing identified key characteristics where SPC can be shown to contribute
little or nothing of value to the product.
If statistical process control is chosen as the method of control
for the key characteristic in 4.2.2 above, the following
requirements must be met:
Section 4.2.4 Key characteristic measurements shall be taken
from the current production products, and shall represent the
normal production output.
Control charts are designed to yield an accurate description of the process as
it functions over time. To do this, only data that represents the current
production output should be used. It is usually not necessary to measure and
chart all (100%) of the production output, but products that are sampled must
represent the entire output.
Section 4.2.5 Control limits shall be computed appropriately
and the limits shall reflect the current process output.
Statistical techniques that are used shall follow industry
standard methods.
One of the purposes of statistical control charts is to answer the question
"Has my process changed?" In other words, "Is my process stable and
predictable?" Whether a process is in statistical control is determined directly
from the control chart being used to monitor the key characteristic. All control
charts define statistical limits for the natural (common cause) variation of a
process. These limits are called control limits.
Control limits are calculated from the data in the process. Since these limits
help answer the questions above, they must be computed correctly. Users
should take great care in making control limit calculations, as they are
affected by sampling rates, sample sizes, the way subgroup measurements
