A Program to Supervise Upjohn's Chemistry Labs
Abstract
For my SIP I worked on a computer project for Upjohn - Control Division as a computer science intern. Control Division monitors the quality of all the various chemicals
and drugs that Upjohn produces. After being given a part of a "lot", Control has its numerous laboratories run various tests to be sure that all of the chemical's characteristics are within standards (i.e., that the drug is fit for use.) One
problem that Control Division poses for the rest of the
company is lapse time. For example, if Upjohn manufactures
a "lot" of aspirin, it can't just sell it to the public right
after it finishes making it. FDA requirements, as well as
common sense, state that the aspirin must be tested to make
sure that it is aspirin and not some toxic variant there of.
Thus the 30 or so laboratories in [Quality] Control will each
run a different set of tests on a sample of the lot. The
"lapse time" is the time between the date at which the drug
is made and the date at which the drug passes inspection and
can be sold. Each lab approves the drug, therefore each lab
must finish its tests before the lot is approved. Thus while
the average time for a lab to complete its tests may be very
low, (one or two days), the actual lapse time on a lot will
be very long if one lab takes a long time (say 30 days) even if all the others take only one day. So, quite commonly, one
lab holds up the entire approval process and while everyone
is waiting, the lot is sitting in a warehouse somewhere
earning no money.
My project addresses this problem by gathering statistics
on all the labs to see which labs are slowing up the testing
process, and thus where efforts need to be made or who needs
to be yelled at. It does no good what so ever to half the
average time taken per lab if the last lab is still taking the
same amount of time. Likewise, a one day improvement by that
slowest lab moves up the sale date of a product by a day.
The former program that did this was so user unfriendly
that only John Nadelin, my supervisor, could use it. It was
inefficient, slow, and didn't produce accurate data as well.
My project fixed all of these problems. It produced accurate
data and was so user friendly that even a manager could run
it.