July 2006

"Quality Focus" - monthly Genesis column

The Canmaking Line's Equation

The process of adding value to a can on a manufacturing line is linear, but the interactions between each process are not...

A mathematical equation – more accurately in this case a statistical equation - exists in every can plant in the world, whether two or three-piece.  Actually all mass production processes are inhabited by such, albeit in varying degrees of complexity.  Suffice to say the one governing can lines is very (very) complex. 

This massive statistical equation is comprised of many systems and sub-systems that propel a can to its finished state.  This series of production events is linear, in that the value-added steps in the line are serial in nature.

However, there are myriads of statistical events outside this linear process, which interact with one another to varying degrees.  Additionally, each process in the line contains its own internal inputs, which interact with each other and with other inputs in other line processes.  In such cases Statistics is the only method viable to quantify these interactions and use them to the benefit of the production process or to solve problems when things go awry.

As an example, basecoated cans – using two-piece cans to illustrate a point – receive their value-added basecoating at the basecoater, a piece of equipment located within the “linearity” of the can line.  The coverage on the can, besides contributing to container aesthetics, affects other processes downstream (decoration, for instance).  But the basecoater is also creating its own potential problems due to the variation being dealt it from within, such as basecoating properties, gravure roller design (and degradation over time), applicator roller condition and degradation, temperature variation, prespin deterioration, mandrel variation and degradation, just to name a few.

The finished condition of the basecoated can is also dependent upon inputs from upstream in the process, such as those created by the washer, trimmer, bodymaker, cupper, lubricator, and all the way back to the uncoiler.  All these processes also have their own internal variation that contributes to the state of the can they produce as well as interacting with basecoating inputs.

The power of unveiling the Statistics behind a can line is the inherent ability to arithmetically deal with the variation.  However, true variation may only be first determined through Statistics, as Statistical variance may be arithmetically manipulated.  Conversely, specifications – or tolerances – should never be.  Why not?  There is a common mass production misconception based upon archaic tolerance stacking assumptions in the engineering world.  To examine a series of parameters and assume some will operate at their high limit and some at their low is incorrect, as statistically this almost never happens.  What actually happens is they all operate at individual means (averages) – and individual sigmas (statistical standard deviation).  From this data a wealth of information may be gleaned to allow for the control of the process, as well as predictive indicators of when the process will go out of control – before it happens.

In our basecoater example, we should not wait and see (inspect) what the basecoater is feeding the downstream processes, we would already know before it happens via knowledge of the input variation to the basecoater.  We can keep the basecoating process under control not necessarily by inspecting its output, but by already knowing its input variation and effects upon our desired output.

Rest assured that quantifying inputs is much easier than measuring – and reacting to – inspected outputs, especially when trying to measure a statistical system (can line) with Arithmetic.  This is not new, as the best-in-class manufacturers (of all products) around the world have learned this and use it to rise far above the competition.