Chapter 13: Short-Run SPC

This chapter presents a special application of SPC for processes that produce limited output, small production runs, and/or small batches. Because these "small" runs are very limited in producing a significant level of variation, short-run SPC is a way to circumvent the variation issue and produce a chart that can indeed separate normal from special variation.

OVERVIEW

The manufacturing environment has undergone many significant changes since Dr. Walter Shewhart first developed control charts at Bell Labs back in 1923. For example, the increasing use of just-in-time (JIT) inventory systems demands small lot sizes with shorter production runs. This means more frequent machine changeovers or setups to accommodate many different part numbers (model number, color , etc.) with possibly different print specifications, all run across the same process. By their very nature, job shops have also had great difficulty in applying the powerful methods of statistical process control (SPC) to their processes because of limited lot sizes.

These short production runs cause problems when attempting to use the traditional Shewhart charts because there are never enough data to calculate control limits in a timely manner. Usually, the run for a given part number is over before the limits can be calculated and drawn on the chart. This means that operators must wait until after the job is completed before they can discover whether the process was in or out of control.

Even when there are enough data for the first part-number run, if a different part number is scheduled to be run on this equipment, a new chart must be started. Because most job shops have hundreds (if not thousands) of part numbers, a mountain of paperwork is created. Operators waste valuable production time searching for the proper chart, thereby decreasing their efficiency. Maintaining all these separate charts, although valid for each individual part number, is not very effective in evaluating the continuous performance of the equipment over time.

Although traditional charts are not applicable, the concepts of SPC are still applicable ; however, they must be adjusted for smaller samples. The adjustments to these charts allow operators to plot different part numbers on the same chart by the use of a special data transformation that scales the data from different part numbers to one common distribution.

This transformation of the data allows operators to draw one common set of control limits on the chart that applies to all part numbers plotted, thus eliminating the need for hundreds of separate charts. Control limits can be determined sooner ( generally , these control limits are set at ±3 sigma), and because all the data are now plotted chronologically on the same chart, any time- related process changes can be more easily detected .

Because the type of short-run chart needed depends on process conditions ”just like any other process under normal conditions ”in this chapter we will show the reader the applicable transformation formulas without any detailed explanation. Readers interested in more information may wish to refer to Bothe (1991), Pyzdek (1992), Hiller (1969), Bothe and Marvin (1985), and Kane (1988).