Predicting Scrap
“Why is the scrap so high?”
I’ll bet a day doesn’t go by that some foundry
manager doesn’t ask that question. I’ll also bet there isn’t a day when a
foundry manager doesn’t get the answer, “We were running jobs that always have
high scrap.”
It’s a very frustrating exchange for managers. They
really want to chew some butt for the high scrap, but they also realize there
are high scrap jobs.
Interesting side note – I don’t remember ever hearing
a foundryman say, “Oh, we’re just running good jobs.” when complimented on low
scrap.
Everyone working in a foundry for any length of time
knows some jobs have higher scrap than others. Those jobs most likely push the
envelope of the foundry’s capability, but if enough of those jobs run at one
time, higher scrap totals should be expected.
Foundry scrap depends upon the processes affecting
every job and the idiosyncrasies of individual jobs. If the processing allows
the metal to get too cold periodically, castings will misrun even if the gating
is well done. On the other hand, if the ingates are too small on a particular
job, misruns happen even if the metal temperature is normal.
What is a manager to do when given the reason for
high scrap as running tough jobs? He
could merely assume he was being told the truth. He would then most likely be left to shrugging
his shoulders and saying something like, "It seems like we're always
running high scrap jobs lately."
Not a very satisfactory response.
A second alternative would be for the manager to
assume that he’s being lied to. His
response then would be something like, "Don't give me that ****! The castings we’ve been running aren’t any
worse than any others!" It’s been
my experience when someone takes this approach, records are usually pulled
showing that jobs that normally do have high scrap were run. The manager is frustrated, and those who were
the recipient of the tirade are further convinced he’s a jerk.
There’s a better way now. Quit looking at total
scrap. Instead, look at the difference
between actual and predicted scrap. With the speed and data handling
capabilities of today’s computers, it’s an easy matter to predict what the
scrap will be based on the history of the jobs run.
Even if scrap is low, if it’s higher than or the same
as predicted, the manager shouldn’t be happy. It’s an indication that effective
steps aren’t being taken to improve the operation. Conversely, if the scrap is
high but below prediction, the manager can take some satisfaction. His people
are improving techniques either on the general processes or on individual jobs.
As Plato said, “Never discourage anyone who continually makes progress, no
matter how slow.”
One should expect the actual scrap would be around
the predicted if nothing has changed; however, it’s been my experience that
isn’t the case. If a foundry doesn’t constantly work on improving, actual scrap
will creep higher. Patterns wear, procedures slip, and scrap rises when effort
isn’t expended.
Many of the scheduling packages currently used in
foundries already take advantage of the logic of predicting scrap. They automatically schedule additional castings
to be poured based on the scrap history of the specific job. Taking that same
information and applying it to all of the jobs and quantities run during a time
period should be an easy programming feat.
While having a program to calculate the prediction
from existing records is the easier way, a simple spreadsheet program can be
used to accomplish the same end.
What to do with data from jobs with no history is a
legitimate question. Predicting the scrap from new jobs can be handled in two
ways. Most operations don’t track the scrap history of new jobs. If that’s the
case, an overall average of all jobs could be used. A better way is to develop
an average for new jobs. Why is that better? It points out a potential
operational problem. If new jobs typically have significantly higher scrap than
existing jobs, the average should draw attention to the need to improve the
process of starting new jobs.
Another benefit of using predicted scrap is that it
draws attention to the specific jobs that do have high scrap. It is those jobs
that deserve the attention necessary to improve their operation. After all,
it’s almost always easier to see substantial improvement when working on a job
that historically has 50% scrap instead of one that has ½ %
Comparing actual scrap to predictions based on previous history doesn’t solve problems for managers. It always takes action to solve the problems, but the comparisons provide an additional tool for managers to determine the direction of their operations.