Re: Short Run SPC

["Rajeev Tatkar" <rajeevtatkar@iname.com>]

We had a problem with short run SPC which we got over through a rather
innovative method.  I do not know if there is anything in theory that can
state things in a manner we did it in practice, nor am I sure that we did
the right thing.  It however worked and hence I thought I will share it with
you.  Here is the problem and what we did:

The problem was that the client manufactured items for which the batch size
was one!  The probability that a part with the same dimensions will ever be
made again was approximately once in six months.  Practically every item
that the client manufactured required a similar part though the dimensions
could vary rather drastically.  How does one apply SPC in a situation like
this?  The method we decided to follow was the following:

If a particular dimension is say 22.5 +0.8 -0.4 we convert the dimension to
100 +10 -10  i.e. 22.5 is 100, 23.8 is 110 and 22.1 is 90.  Similarly if
another dimension is say 63.25 +.05 -0.02 then 63.25 is 100, 63.3 is 110 and
63.23 is 90.  Plot the results that you get and apply all your analytical
skills to determine what is happening to the process.

We were of course initially looking for only recording the variation.  An
analysis showed that there were indeed specific patters that one obtained
when some of the parameters namely, material and men were varied.  Once a
cause had been established the idea was to remove the cause of the variation
and plot further to see if the variation has indeed reduced.  The whole idea
here was to find assignable causes and to remove them.  While the variation
initially was indeed high - with a substantial number of points even falling
outside the control limits, methods were evolved to take care of the
variation and avoid recurrence.

Rajeev Tatkar
rajeevtatkar@iname.com
Smart System
"Show us a business we cannot help you improve"
=========================================================================