Many manufacturers can reduce manufacturing costs and customer satisfaction problems by improving the procedures used to set specifications. Instead, they spend more than required for material and components, or they have substantial problems in production due to parts that don't work together as designed. They may also have customer satisfaction problems with reduced functionality or lifetime of a product. These problems might be substantially reduced through more appropriate tolerancing.

The word "Tolerance" is defined as the range of allowable variability, which is the difference between upper and lower specification limits for a particular product quality characteristic.

Integrated Tolerance Management is defined as a data-driven system for establishing product and process specifications to simultaneously increase customer satisfaction, speed time to market, and reduce costs.

Integrated Tolerance Management requires the collaboration of all the parts of the organization depicted in the above figure. Specifications on product and components, typically set by product design engineers, are chosen to balance the desired with the possible. Information regarding what is desired typically comes from Marketing. Data on the possible -- process capability data -- is sometimes available from Production and Incoming Inspection. For materials that are not used in current products, Purchasing might help obtain relevant data from vendors.

Integrated Tolerance Management helps engineers do a better job of connecting product and component specifications by providing them with better information about both the desired and the possible. For manufacturers that use Statistical Process Control, further improvements can be obtained by specifying limits on the process average and standard deviation and avoiding tolerancing individual parts; this practice is called "process tolerancing".

Process Tolerancing

"Process Tolerancing" is defined as setting specification limits on the process average and standard deviation. Process tolerancing rests on two basic observations. First, the primary use of tolerances in most cases is to control processes and through that to achieve quality. Second, process tolerancing can substantially simplify the procedures used to tie component to assembly specifications.

Specifications are needed on individual parts in only two cases: For some products, they may be required for advertising. In other situations, 100 percent inspection with substantial scrap or rework may be required to achieve the level of process control required in manufacturing. These two cases cover only a small percentage of specifications used in industry. Even in these cases, it is generally easier to assure quality by managing processes than by focussing exclusive on whether individual measurements are within specification limits. In all cases, process management is improved by asking for what you want, namely by specifying limits on process bias and standard deviation.

There has been substantial discussion since 1929 about how to relate specifications on components to specifications on assemblies. So-called "worst case" tolerancing has been criticized because the extreme cases it is designed to handle only rarely occur. An alternative repeatedly recommended since 1929 has been "root sum of squares statistical tolerancing." However, we have recently established that "root sum of squares statistical tolerancing" is extremely non-robust to violations of assumptions. (For a report on this issue, contact the Center for Quality and Productivity Improvement at the University of Wisconsin-Madison and ask for Report 159 on "Five Ways Statistical Tolerancing Can Fail and What to Do About Them" by Spencer Graves and Soren Bisgaard.) Most of these problems can be avoided by placing specification limits on processes instead of (or in addition to) specifications on individual units. Most of the remaining problems can be avoided by using the results of properly designed experiments for tolerancing, as described by Bisgaard, "Designing Experiments for Tolerancing Assembled Products," Technometrics, vol. 39, pp. 142-152.

For more information or for consulting assistance with tolerancing, call (408)294-5779, fax: (408)294-2343, or e-mail: sgraves@prodsyse.com.

Other Productive Systems Engineering programs: Product Development, Designed Experiments (Design of Experiments, DoE or DOX), Reliability Experimentation, Quick Start Total Quality (Total Quality Control, TQC / Total Quality Management, TQM), Statistical Process Control (SPC); Fiber Optic Communication/Transmission Systems (FOCS or FOTS); Control and Monitoring (RMM); Electronics (Theory & Application); Production Line Assembly for Technicians (Assembly).