June 2006

"Quality Focus" - monthly Genesis column

Identifying the Domino Effect

Before identifying the statistical behavior of a canmaking plant, it's vital to pinpoint the variables...

The past five months of this column built a foundation upon which we may begin to specifically discuss variation in a can line.  Although rather basic – given space restrictions in the magazine – this foundation allows us to better understand the statistical “heartbeat” of a can plant.  Now we will begin to examine the practicality of statistically capturing canmaking process variation; as always, suppliers, do not turn the page, as you are heavily involved here. 

Canmaking is an amazing process; however, with this comes a very (very) complicated statistical series of events inflicted upon each and every container.  Just because we may not be measuring them – directly or indirectly – does not mean they do not exist: we simply may be (unintentionally) avoiding them.  It is somewhat analogous to visiting one’s physician: we may want to know the state of our health, but sometimes we are content with thinking we need not fix something that we perceive as “not broken.” 

From a mathematical – and specifically statistical – viewpoint, a can line is a very large and complex “equation;” yes, everything that happens in production can be somehow quantified via application of almost all of the sciences.  This process starts –for purposes of our discussion – at the coil (or sheet).  For illustration we will focus our attention upon two-piece can manufacture; three-piece is similar enough to draw significant parallels. 

And for purposes of this discussion, the end of the process is final “inspection” of the containers.  What happens in between these two points is mathematically astounding. 

Think about the coil for a moment.  Is there any variation here?  Those of you who have switched coil types on can lines know that sometimes these variations – in this case between two substrate manufacturers – can be very significant.  The variation within a single coil is simply a microcosm of the larger picture; it exists, but the issue is its severity and resulting effects upon subsequent operations and ultimately upon the finished product. 

For example, what properties of the coil can vary?  Thickness?  Material properties?  Grain direction?  Surface energy?  In turn, how does the variation in the Uncoiler affect the delivery of the coil to the lubricator and cupping press?  Keep in mind here that we are “picking on” the coil to start, only because we have defined it here as the first piece of our puzzle; we will eventually get to other can line components in subsequent columns. 

What about the Lubricator?  What is the variation in dispersion across the coil?  Along the coil?  And the lube, for that matter?  How does the variation in its composition affect cupper tooling?  Cup “quality?”  Cup variation? 

It has been my experience that these type factors are generally understood – as they should be – but usually expressed as “being within specification,” or “within some upper and lower range,” etc.  The problem is that with this type measure there is little, if anything, we may infer regarding the statistical capability of the line.  Each of these factors – and here we have limited ourselves to a scant few to start – are what need to be measured, not how often we are “in spec” or how many “defects” we have made.  This is 20^th century thinking that will no longer work in the long term. 

The issue is that each key input factor in a can line compounds the process variability.  Have you – canmaker and supplier alike – ever been faced with a problem that seemed irresolvable?  Keeps coming back even after “fixes?”  Never went away?  Chances are true statistical methods – versus “pseudo-statistics” – were never used.  Yes, it is that simple to explain; stay tuned for more.