Control Charts

Manufacturing operations of technology companies employ complex processes that need to be under strict control at all times.  Controlling a process usually means maintaining its output parameters to within certain specifications by providing it with the correct set of inputs.  Unfortunately, keeping the process inputs within their defined specifications is not often enough to ensure that the output will always be good. Other factors that were not initially considered in the design of the process can come into play and degrade the performance of the process, even if the inputs to the process follow the specifications. 

Control charting is a technique for monitoring the performance of the process for any signs of deterioration, so that actions may be taken before the process gets out of control. This technique consists of plotting critical process output parameters on control charts at defined intervals, and analyzing the trends exhibited by the plots for any abnormalities that need intervention.

A control chart is used:  1) for presenting process performance in a quick and easy-to-use visual format; 2) for monitoring process variation over time; 3) for distinguishing out-of-control points due to special assignable causes from variations due to common causes that are part of the process;  4) for detecting abnormal trends and other tell-tale signs of process anomaly; 5) as feedback for processes that are undergoing improvements; and 6) as a common language for discussing process performance. 

Control charting can not be applied to every process though.  It can only be implemented for processes that are already stable, and whose output data for charting constitute a normal distribution.  A stable process is one whose output data form a distribution with low variation, i.e., the data have a low standard deviation. On the other hand, an unstable process exhibits very large variation, i.e., the output data have a high standard deviation.

The high variation of an immature or unstable process is usually due to a number of extremely high or extremely low points (known as out-of-control points or outliers) caused by special random causes that are not part of the process itself. Such factors must be minimized (if not eliminated) first, to yield output data that truly reflect the inherent capability of the process, before control charting is started.  Doing so will ensure that the process under control charting exhibits variation caused only by factors that are part of the process itself.

There are many types of control charts for both attribute and variable types of data.  However, the control chart used for individual readings of variable data will be used in the following discussions since it is one of the most extensively  used control charts in process engineering.

A completed control chart has the following parts: 1) an x-axis that shows the points at which the parameter readings were collected; 2) a y-axis that shows the parameter reading for each data collection point on the x-axis; 3) a horizontal data average or process mean line; 4) horizontal line(s) for the control limit(s); 5) horizontal line(s) for the specification limit(s); 5) a horizontal target line lying exactly between the specification limit lines; and 6) the plotted data.  Figure 1 shows an example of a completed control chart.

Figure 1. Example of a complete control chart

<Proceed to Page 2>