Controlling processes statistically.

CONTROLLING PROCESSES STATISTICALLY

Surprisingly, perhaps, SPC has its roots in the work of Dr. Walter Shewhart at the Bell Laboratories dating right back to the 1920's. Shewhart viewed a production process as an integrated system with elements consisting of people, equipment, materials, methods and environment.

He noted that every process has some average output and some random variation around that average. The randomness is unavoidable and stems from all the possible causes and effects that are not under control. Such an output could be depicted by a histogram, and its related frequency distribution curve.

While many processes may produce an approximately normal output distribution, many do not. Shewhart, however, would periodically take sample readings of the process and plot the sample averages. When sample averages are plotted, they always conform to a roughly normal distribution even if the parent distribution is anything but normal. This is known as the Central Limited Theorem in statistics.

Frequency distribution curves can vary as to mean value, spread and shape. If the process remains in statistical control, the random distribution of output over time will be predictable and repeatable, noting that each of the distribution curves shown might represent one hour's, one shift's, or one day's worth of output. When, however, a process is not in statistical control, the output distribution can vary as to mean, …