May 23 2005

Changing CpK for Profit

Published by at 5:15 pm under Uncategorized

We go to a significant amount of trouble to qualify a tool and maintain sometimes, some unreasonably high CpK values.  Over time the process degrades, the tooling wears BUT the CpK never changes.  The purpose of Process Capability is to assure that good parts will be produced in a robust manner.  However it never says some engineer’s or quality control specifications are chiseled in stone for fear of bad products.  A Process Capability of 1.0 or greater assures the parts will be within the print specification.  This can be used to improve productivity and profits.Cp and CpK are usually tools used in qualifications.  Cp (Process Capability) simply answers the question: “If your tolerance is +/- .003 Inches, can any of the cavities be within this tolerance?”  Interesting question.  It’s like saying with a bow and arrow you can place a dozen arrows within a 12-inch group BUT it never asks the question if those arrows were even aimed at a target a yard in diameter!  Opps.  However a unilateral process tolerance (CpK), asks both if they are grouped tight enough to meet the specified tolerance AND if they are within the specified tolerance (read: “is your grouping tight enough and were you polite enough to aim at it AND hit?”).


The CpK formula says to take the ‘distance’ from the center of the tolerance, add in three standard deviation units, and then divide this into half the tolerance.  This tells you the probability of drifting out of print.  If everything fits in nicely, this number is above 1.0.  If there’s a chance of producing out-of-print parts, the number will be less than 1.0.


The Quality Vultures and the Data Dummies only come out to the production floor when the mold is all bright and shiny for the approval run.  They measure, calculate, measure again, re-calculate, draw a few charts, have a meeting, do a kissy-kissy PowerPoint presentation for upper management, and then declare the mold capable for production.  They forget the CpK value they got is a function of usually one very specific set of process conditions at the time of the trial.  But so long as the number is above 1.0 (dare we say it? Even if it changes but NEVER goes below 1.0) the parts will be statistically ‘good’.  This technique can be exploited.


Months later the Vultures and Dummies are calculating calculations, presenting presentations, and still kissing managers.  You are on the midnight shift staring at the parts asking yourself why the tech crew slowed everything down to achieve this CpK because these conditions are now considered Holy Writ.  They cannot be changed even if there is an opportunity to improve productivity, profits, and a nice comment on your performance review.


Telling anyone in middle/upper management the process can be improved is like telling a 5 year old there is no Santa Claus.  They’ve seen the approval data and don’t want to see anything else.  They like their charts, graphs, and multi-colored simulations/wallpaper.  Asking to do an independent study is a blasphemy.  What’s an engineer to do?


The best technique is to “manage from the bottom.”  Your job is not to tell the boss he’s wrong, but get him to a place where he’ll tell you the numbers are goofy and maybe he can approve a quickie study and prove “He’s” right.  Since it will be ‘his’ idea, he can then “manage from the top” telling you to do your study on a small insignificant part with little financial impact.  You do this by pointing out that it will take a short period of time (a few hours), at worst it will produce a little scrap (you do anyway), and if it goes into Disaster Mode you can quickly put it back to the original settings.  Your anticipated outcome will be a reduction in cycle time at no cost to quality therefore improving profit, productivity, opening up machine time, and making him the “flavor of the month” when all the excess profits come rolling in. You also point out the data you come up with will make a snazzy PowerPoint presentation where he can show to Quality, The Data Dudes, Other Managers and Upper Management that HE took the initiative and improved the bottom line using side-by-side data.


OK – Here’s how you do the study:  You already know you have a given acceptable CpK, productivity rate, etc. etc.  Most technicians set (1) the packing time too long and/or (2) the cooling time too long. Because it is safe. You can only pack until the gating system if frozen.  This is a simple study in weighing parts, looking for a weight loss as you shorten the packing time.  As soon as the weight loss is noticeable, increase the packing time slightly until you have a stable weight and a slight safety margin of time.  Excess cooling time uses the mold as an unintended shrink fixture.  Unless the cooling time is intentionally long, taking a second off the cooling time will usually not affect anything.  Removing unneeded packing time is the same as cooling time.  Proceed slowly and carefully. Dimensions will change.  This can also be done with mold/melt temperatures, the platen open and closing distances and/or any other parameter you might think of.  Sometimes a DOE will point you in the right direction.


With your new settings, let the process equilibrate for about 20+ minutes, and then take your samples – between 20 and 30 shots.  Sometimes (just for giggles) you might want to try a series of different combinations of settings gathering several samples.  Just make sure the process doesn’t drift after several shots.


CpK is calculated on one dimension, thus the measurements and calculations are fast and easy.  Now Logic and Sanity must be used:  If the original CpK was 2.3 (which is very tight) and your test samples lower it to 1.4, all you have done is opened up either the spread of the samples or the position with respect to the spec’s center.  NEVERTHELESS. . . . .  The parts are still statistically within the specification and to print.  Bingo!  You’ve improved productivity and profits with no increase in rejects.


Gather up your data, calculate the savings/profits (plus executive bonus’s etc.), and toss it off on the boss for PowerPointing.  This is called a Shining Star experiment.  One little study with a small downside but a demonstratable upside can be leveraged to other parts, making for a big upside.


This entire procedure would be silly and redundant if the process had been optimized in the first place before people began to generate the statistics.  Yet, most Techs and Quality people have bought into conservative production is better than productive production because it is a form of job security.  This leaves a lot of profit left on the table.  It only has to be picked up.  If your boss looks good, that’s politics.  If you remind him at your next performance review, it was your efforts that are continuing to make him look good; it’s a nice bargaining chip to have to cash in for a raise because you’ll look like you are Head and Shoulders above everyone else. 


Here are the rules:

Ø       Take the LOWEST CpK possible.  Your interest is in parts that work, NOT part to print.

Ø       Optimize the process, don’t mandate the CpK. – Process capabilities are the combination of Real World Measurements and Designer Inspired Wishes.  The process will give you what it will give you and nothing more.

Ø    Remember things change.  During your qualification run, look for the widest possible processing window that will still give you a CpK larger than 1.0.  This will keep productivity up.

Bill Tobin

No responses yet

Trackback URI | Comments RSS

Leave a Reply