There is some disagreement among quality professionals whether or not precontrol is a form of statistical process control (SPC). Like many tools prescribed by the Shainin System, precontrol’s statistical sophistication is disguised by its simplicity. The attitude of many seems to be that if it isn’t difficult or complex, it must not be rigorous.
Despite its simplicity, precontrol provides an effective means of process monitoring with several advantages (compared to control charting), including:
Successful implementation of precontrol begins with an evaluation of the process to be monitored to ensure that it is a suitable application. Information about existing processes should be readily available for this purpose. New processes should be carefully considered, with comparisons to similar operations, to determine suitability. Precontrol is best-suited to processes with high capability (i.e. low variability) and stability (i.e. output drifts slowly).
Operators’ process knowledge is critical to the success of precontrol. They must understand the input-output relationship being monitored and be capable of making appropriate adjustments when needed. Otherwise, their interventions are merely process “tampering” that results in higher variability and lower overall performance. Reliable measurement systems (see Vol. IV, Vol. V) are required to support effective process control by operators without excessive intervention.
Process Monitoring Zones
Process monitoring zones, or precontrol zones, are based on the tolerance range of the process output, with the target value centered between the upper and lower specification limits (USL, LSL) of a bilateral (two-sided) tolerance. The relationship of this tolerance range to input variables should be known in order to make effective process adjustments when needed. Tolerance parallelograms, or other techniques, can be used for this purpose.
To define the precontrol zones for a bilateral tolerance, divide the tolerance range into four equal segments. The two segments that flank the target value are combined to create the green zone; measurements that fall in this zone are acceptable. In normally-distributed data, this 50% of tolerance encompasses approximately 86% of process output.
The remaining segments, each containing 25% of the tolerance, are the yellow zones. The yellow zones are often called warning zones, because measurements that fall in these zones may indicate that the process has drifted and requires adjustment. In normally-distributed data, approximately 7% of process output will fall in each of the yellow zones. These estimates assume that the width of the normal distribution matches the tolerance range and its mean equals the target value.
The precontrol red zones encompass all values outside the specification limits. A graphical representation of bilateral tolerance precontrol zones is shown in Exhibit 1.
There are three possible precontrol zone configurations for unilateral (one-sided) tolerances. The first, called “zero is best,” simply divides the tolerance range by two. The green zone encompasses the 50% of tolerance nearest zero (lower half) and a single yellow zone encompasses the remaining tolerance, up to the USL (upper half). A single red zone encompasses all values above the USL. This configuration is used for measurements that cannot produce negative values, such as surface roughness or yield loss. “Zero is best” precontrol zones are shown graphically in Exhibit 2.
The remaining two configurations are, essentially, mirror images of each other. In one case, the LSL is defined, with no upper bound specified (“more is better”); the other defines the USL, while no lower bound is specified (“less is better”). For each case, the tolerance range used to define precontrol zones is the difference between the specification limit (LSL or USL) and the best output that can be expected from the process (highest or lowest). A single yellow zone encompasses 25% of the “tolerance” nearest the specification limit.
The green zone includes the remaining 75% of “tolerance” and beyond. Any measurements beyond the “best case” value should be investigated. The expectations of the process may require adjustment, leading to recalculation of the precontrol limits. It could also lead to the discovery of a measurement system failure. Precontrol limits should also be reviewed as part of any process improvement project.
The red zone in each case encompasses all values beyond the specification limit (y < LSL or y > USL). “More is better” and “less is better” configurations of unilateral tolerance precontrol are presented graphically in Exhibit 3.
Defined precontrol zones establish the framework within which process performance is evaluated. The remaining component defines how to conduct such evaluations via decision rules that guide setup validation or qualification, run-time evaluations, and sampling frequency. These decision rule sets are presented in the following section.
The first set of decision rules define the setup qualification process. To approve a process for release to production, the measurements of five consecutive units must be in the green zone. The setup qualification guidelines are as follows:
Once the process has been released to production, a new set of decision rules are followed. For run-time evaluations, periodic samples of two consecutive units are measured. Response to the sample measurement results is according to the following guidelines:
The final decision rule defines the sampling frequency or interval. The interval between samples can be defined in terms of time or quantity of output. The target sampling interval is one-sixth the interval between process adjustments, on average. Stated another way, the goal is to sample six times between required adjustments. For example, a process that, on average, produces 60 units per hour and runs for three hours between required adjustments should be sampled once every 30 minutes or once per 30 units of production. The sampling frequency may change over time as a result of learning curve effects, improvement projects, changes in equipment reliability, or other factors that influence process performance.
The simplicity of precontrol, demonstrated by the division of the tolerance range into precontrol zones and easily-applied decision rules, makes it an attractive tool for implementation in production departments. Administration of such a tool by those responsible for production maximizes its utility; it takes advantage of the expertise of process managers and operators and eliminates delays in response to signs of trouble in the process.
Modifications to Precontrol
The formulation presented above may be called “classical precontrol;” it serves as the baseline system to which modifications can be made. F. E. Satterthwaite’s original formulation (1954) prescribed a green zone containing 48% of the tolerance range and yellow zones containing 26% of the tolerance range each. The 50%/25% convention was adopted for ease of recall and calculation in an era preceding electronic aids. If such aids are in use, the choice of zone sizes is nearly imperceptible in practice. Either scheme can be chosen, but it should be used consistently throughout an organization to avoid confusion.
Two-stage precontrol retains the precontrol zone definitions and the setup qualification and sampling frequency rules of classical precontrol, but expands the run-time evaluation rules. In two-stage precontrol, responses to sample measurement results are in accordance with the following guidelines:
Modified precontrol is a hybrid of classical and two-stage precontrol and control charts. The setup qualification and sampling frequency rules of classical precontrol are retained. The run-time evaluation rules are the same as those used in two-stage precontrol. Precontrol zone definitions are adapted from Shewhart’s control limits. Green zone boundaries are defined by ±1.5σ (standard deviations of process performance), while yellow zones occupy the remaining tolerance range (±3σ). This version negates one of the key advantages of classical precontrol – namely, no calculations required for evaluation – but the resulting sensitivity may be needed in some circumstances. With increased sensitivity, however, comes a higher rate of false alarms (Type I error) that prompt adjustments that may be unnecessary.
While other modification schemes exist, a thorough treatment is not the objective of this presentation. If one of the formulations outlined above does not suit your needs, the presentation should suffice as an introduction to possible modifications. To find a more suitable process control method, the cited references, or other sources, can be used for further research.
As stated at the outset, charts are not required to implement any of the formulations of precontrol described above. However, a precontrol chart can be a useful addition to the basic tool. Charting provides historical data that can be applied to process improvement efforts or to detect excessively frequent adjustments, called tampering. A precontrol chart can also provide an indication of operators’ effectiveness in making adjustments or the need for additional training.
The final note to be made is less a modification than a recommendation. The previous discussion of precontrol has been based on its application to process outputs. While this is useful, the power of precontrol is maximized when it is applied to process inputs whenever practical. This proactive approach can prevent high input variability from negatively effecting process output, reducing the number of samples, stoppages, and adjustments required to produce the demanded quantity of output.
Despite its advantages, precontrol seems to incense some vocal advocates of SPC and control charting. The criticisms of precontrol are not discussed her in detail for the following reasons:
The decision to implement precontrol, in any configuration, or any of the alternatives, requires careful consideration of the process, operators, and customers. Implementing an inappropriate or ineffective process control method can damage the credibility of all future efforts. This may lead to process tampering or, on the opposite end of the spectrum, neglect. The application must be monitored to ensure the system supports the ongoing effectiveness of operators in maintaining required quality and productivity levels.
Contact JayWink Solutions for assistance in evaluating processes, establishing a precontrol system, training, or other process monitoring, control, and improvement needs.
For a directory of “The War on Error” volumes on “The Third Degree,” see “Vol. I: Welcome to the Army.”
[Link] “Strategies for Technical Problem Solving.” Richard D. Shainin; Quality Engineering, 1993.
[Link] “An Overview of the Shainin SystemTM for Quality Improvement.” Stefan H. Steiner, R. Jock MacKay, and John S. Ramberg; Quality Engineering, 2008.
[Link] “Precontrol.” Wikilean.
[Link] “Pre-Control: No Substitute for Statistical Process Control.” Steven Wachs; WinSPC.com.
[Link] “The Power of PRE-Control.” Hemant P. Urdhwareshe; Symphony Technologies.
[Link] “Pre-Control May be the Solution.” Jim L. Smith; Quality Magazine, September 2, 2009.
[Link] “Using Control Charts or Pre-control Charts.” Carl Berardinelli; iSixSigma.
[Link] “The theory of ‘Pre-Control’: a serious method or a colourful naivity?” N. Logothetis; Total Quality Management, Vol 1, No 2, 1990.
[Link] “Precontrol.” Beverly Daniels and Tim Cowie; IDEXX Laboratories, 2008.
[Link] “Shewhart Charts & Pre-Control: Rivals or Teammates?” Tripp Martin; ASQC Statistics Division Newsletter, Vol 13, No 3, 1992.
[Link] “Pre-control and Some Simple Alternatives.” Stefan H. Steiner; Quality Engineering, 1997.
[Link] “Pre-control versus and R Charting: Continuous or Immediate Improvement?” Dorian Shainin and Peter Shainin; Quality Engineering, 1989.
[Link] World Class Quality. Keki R. Bhote; American Management Association, 1991.
Jody W. Phelps, MSc, PMP®, MBA
JayWink Solutions, LLC
If you'd like to contribute to this blog, please email firstname.lastname@example.org with your suggestions.
© JayWink Solutions, LLC