Six Sigma SPC - Statistical Process Control

Design Margin - Statistical Tolerance
Part of DFSS - Design for Six Sigma

Six Sigma SPC - 2070 W. Washington St. #5 Springfield IL 62702  Ph: 217.698.0063
6 Sigma Statistical Process Control (SPC) Software for Windows 95/98/NT/2000/XP/ME

Home Search Capability Studies Links
Products Services
Prices FAQ
About Support
Contact Quality Control Dictionary and Glossary Free Desktop 6 Sigma Calculator Forum
Quality Books Six Sigma and SPC Articles Free Online 6 Sigma Calculator Free Newsletter
These articles are from the Six Sigma SPC Newsletter and other publications. All articles written by Jim Winings
From various 2003 Newsletters

Design Margin - Statistical Tolerance Part 1
for use with DFSS - Design for Six Sigma

You cannot expect to get to six sigma, (for detail information why, see our six sigma page), on if your specification limit and/or manufacturing process was based around a 3-sigma design margin. If you manufacturing equipment has a tolerance of 3 sigma, there isn’t a whole lot you can do about it except change your calculations by doing estimates, or change the specifications. The latter is the only real way to get to a six sigma design margin. You need to know what your actually process is capable of. Then compare that to this table as a first step. You do this by using control charts that look at X Bar, Range, and a distribution of how your process looks.

If we were to describe the meaning of +/-6 Sigma and Processes, we might describe it as the most wanted attributes that can be used 99.9997%, (3.4 PPM), of the time in an end product and any process that will not fabricate any damage to the end product in excess of 3.4 PPM. Note the saying “as the most wanted attributes”. You do not and probably cannot get 3.4 PPM on every parameter for every piece part, sub-component or assembly that you manufacture it depends on the parameter design. Trying maybe just a big waste of time and could directly contribute to many company’s failures, (sometimes caused by external failures), to introduce a six sigma program.

The product design can and does contribute to product failures. The design of the product, to include the supplier’s designs, is the major determinant of a persistent failure rate, which must be discovered and corrected. Lack of operating margin, (design margin), or lack of an adequate amount of overstress tests tends to appear as a continuous failure rate in the long term.

Reduction of defects levels may be accomplished through these steps.

  • Ascertain the product characteristics that are crucial to satisfying both physical and functional requirements of the consumer, in other words, the customer requirements.
  • Ascertain the exact product components that directly contribute to accomplishing these crucial characteristics.
  • For each ascertained product component establish the process step(s) or process choices that affect or control the required characteristics.
  • Ascertain the upper and lower allowable tolerance, (real), for each of the product(s) characteristics and process(s) steps, which will insure the absolute best performance.
  • Ascertain the product capability part(s) and processes components that will control the absolute best performance.
  • If Cpk, (process capability), is not equal to or greater than 2, then make the necessary adjustments to the product and/or process to ascertain it.

Some points to consider when designing a new product or processes and retrofitting a legacy product or processes to aid with the above steps.

  • Use the fewest number of piece parts, sub-components or assemblies. Mathematically, the fewer points than can go wrong will go wrong.
  • Use parts and/or designs with a know competence that have been proven.
  • Create designs for maximum intolerance of parts.
  • Use the lowest possible stress levels. Do not design for any piece parts, sub-components or assemblies to operate at the edge of their envelope.
  • Know and supply maximum likely operating margins.

When you design a process and product to be manufactured from the beginning using the above guidelines, you can improve reliability due to…

  • Lack of early life failures determines the customer’s perception of the product and the company.
  • Latent defects are reduced, thus reducing improving first time yields. Latent defects are defects that are not easily found by normal test and inspection techniques.
  • No test or screen can find all defects. If there are none to be found, it is much more desirable. Only integrated designs for manufacturability and tight process control can eradicate latent defects so there are none to be found.

If you have piece parts, sub-components or assemblies that need repair after the product has started the manufacturing process, the cost goes up and the latent defect rate also rises because of the need to handle the product more. This can cause additional manufacturing errors. Solder shorts, product may be dropped in the repair process causing defects that have already been test or inspected for. It also forces you maintain a larger inventory, thus, inventory turns are larger. An inventory turn is the amount of time it takes for a facility to turn it’s entire stock of inventory and be replaced. Example, a turn of 1.25 could mean 288 turn days, a turn of 40 could mean 9 turn days. Maintaining a large inventory can get rather costly in terms of dollars that are not accessible. Smaller inventories give you those dollars back for operating capital, etc.

A lot of manufactured products these days are non-repairable. This means that if you do not get it right the first, you have all scrap. But, it’s not just the cost of the scrap, you also have labor involved in producing the non-useable material. And of course the more points of test or inspection the product has gone through before it is detected, if it is detected, adds exponentially to all costs.

While there are formulas to calculate all these items, it is beyond the scope of this document to provide them.

In summary, defects cost lots of dollars due to…

  • The need for additional analytical time
  • The need for additional repair time
  • The need for additional retest time
  • The need for additional re-inspection time
  • The need for additional inventory
  • The need for additional labor and associated costs (benefits)
  • The need for additional materials
  • The need for additional capacity (production equipment / floor space)
  • The need for additional support
  • The need for additional warranty repairs
  • The need for additional replacement of escaping defects

Defects and product design margins have a major role in early life failures, cycle time, inventory and resources and these are key components in competitive issues, which can improve your companies continued existence.

Our Six Sigma software ZeroRejects can help you in many of these areas. Because our histogram chart not only shows you the mean and +/- 3 sigma of the process you are measuring, but also the nominal and +/- 3 sigma of your specification, arranged in units of sigma. This allows you to easily reckon what you specification should be to meet any desired design margin.

Because it is so simple to use, labor can virtually run it with little to no training, this not only saves training time with the initial use, but with every use after t hat, even if there is a month between uses.

Between the initial price, $150.00(US) and updates of $50.00(US), and with the lack or required training time, you can get a return on the investment in less than a few months. Not many statistics programs allow for that. And that is our business plan, for our customers to be able to get a RONI, (return on net investment),  in the smallest amount of time possible. I can put in a sample of 5 data points in less than 20 seconds because it was designed that way. Please take a look at our features page for additional information. We think you will agrees that $150.00(US) is a small price to pay for what this tool does, even for the seasoned black belt, or CQE.

Continue to Part 2

* In special cases where the dimensions do not combine linearly, or are not independent, more complicated calculations may be needed to get the final dimension and the standard deviation.

1 Reference AT&T Statistical Quality Control Handbook Copyright 1956 Western Electric Co., Inc.



Copyright © 2005 Six Sigma SPC / Jim Winings All Rights Reserved
Privacy Statement - Disclaimer - Copyright - Site Map

Last Updated: Sunday, 11-Jun-06 07:19:10 PDT