Page 172 - The Handbook for Quality Management a Complete Guide to Operational Excellence

158 P r o c e s s C o n t r o l Q u a n t i f y i n g P r o c e s s Va r i a t i o n 159

then inspects a sample from the lot or batch and, based on the results of the
inspection, they determine the acceptability of the lot or batch. Acceptance
sampling schemes generally consist of three elements:

• The sampling plan. How many units should be inspected? What is
the acceptance criteria?
• The action to be taken on the current lot or batch. Actions include
accept, sort, scrap, rework, downgrade, return to vendor, etc.
• Action to be taken in the future. Future action includes such
options as switching to reduced or tightened sampling, switching
to 100 percent inspec tion, shutting down the process, etc.
Acceptance sampling methods are generally based on ANSI/ASQ Z1.4
(formerly MIL-STD 105E), or variants of the plan known as Dodge-Romig
Sampling tables.
These acceptance sampling plans have absolutely no place in a modern
quality organization. They should be soundly rejected by the quality profes-
sional. The sampling plans are fundamentally flawed in assuming the sam-
ples are from a homogenous population (i.e., characterized by a single
statistical distribution), when there is no evidence that the samples have been
drawn from a stable process (the only situation under which the samples will
be from a single distribution). When applied to an unstable process, the reli-
ability of the acceptance sampling plan is misstated, and in fact unpredictable
(since, by definition, the output of an unstable process is unpredictable).
If a process is in continuous production, Deming (1986) showed it is
better to use a p chart (a control chart discussed in the next section) for
process control than to apply an acceptance sampling plan. Based on the
stable p chart you can determine the process average fraction defective,
from which you can determine whether to sort the output or ship it by
applying Deming’s all-or-none rule:

If p < K /K then ship, otherwise sort
1 2
where K is the cost of inspecting one piece and K is the cost of shipping
1
2
a defective, including lost customer goodwill. For example, if K = $1 and
1
K = $100 then output from a process with an average fraction defective of
2
less than 1 percent would be shipped without additional inspection; if
the process aver age were 1 percent or greater, the output would be sorted.
Note that this discussion does not apply to critical defects or critical
defec tives. Sampling for critical defects or defectives is done only to con-
firm that a previous 100 percent inspection or test was effective.
As Deming’s rule shows, the alternatives are no inspection, 100 per-
cent inspection, or sampling and analysis using a statistical process con-
trol chart. Supplier programs must emphasize the need for process control
as a condition of sale, as discussed later in the chapter.

09_Pyzdek_Ch09_p151-208.indd 159 11/21/12 1:42 AM