MEASUREMENT ERROR CONCEPTS

W. A. Shewhart has been quoted many times as saying, "In any program of control we must start with observed data; yet data may be either good, bad, or indifferent. Of what value is the theory of control if the observed data going into that theory are bad? That is the question raised again and again by the practical man."

Unfortunately, the analysis of the measurement process within any typical industrial, or other, environment is either ignored or only sporadically conducted . In point of fact, many industrial personnel freely admit their faith in the capability, accuracy, and precision of their gaging as a direct function of the cost of the gage. The general feeling seems to be that if the gage costs more than $80 “100,000 (or perhaps has a digital readout), it must be OK.

Sadly, this assertion is frequently far from the truth. Compounding the problem is that, in addition to the misconception that money can buy gage capability, the basic concept of measurement error, and its relationship to precision and accuracy, is generally misunderstood. Further, the literature available on this subject is highly confusing and often reflects disagreement as to the basic principles of measurement. Juran and Gryna (1980) take special note of this problem, observing, "There is much confusion on terminology. Some literature uses the terms accuracy and precision interchangeably.... The confusion extends to instrument catalogs."

Because of the importance of the topic of gage capability as reflected by Shewhart's concern for the collection of reliable and valid data, the misconceptions and confusion related to the topic reflected in the literature, and the need for a well-defined set of basic procedures to be used by plant and supplier personnel for this effort, this chapter has been prepared. In its development, some decisions had to be made regarding the selection from the several so-called standard definitions of the meaning of the terms included in this chapter. Invariably, these definitions will not be commensurate with all of the definitions and descriptions of those same terms that are used in some of the literature or by some industrial personnel. This, because of the nature of the literature and the topic itself, was unavoidable. The basic definitions for the description of precision and accuracy have been drawn from Juran and Gryna (1980) and the Automotive Industry Action Group (AIAG) MSA (1995, 2002). Many of the remaining definitions were developed from Speitel (1982). These definitions will be presented in the next section.

To start our discussion in measurement, let us begin with the basic applications for the methods and procedures outlined in this chapter, which are

• Evaluation of new gaging

• Comparison of one or more measuring instruments

• Evaluation of equipment vs. gage variability

• Comparison of measuring equipment before and after repair or adjustment

• Determination of true process capability by accounting for the presence and level of variability due to measurement error

The activity of measurement, that is, the acquisition of data using a gage or measuring instrument of some type, may itself be thought of as a process (see Figure 15.2). As reflected by this illustration, the measuring activity ( variables data) may be considered a process, the products (output) of which are discriminating values, observations, or numbers . As with any other operation process, the product is subject to variability. For the moment, let us think of this variability in the context of repeated measurements of the same item or product that yield a series of nonidentical values. The benefit of considering measurement as a process becomes obvious.

Variability in measures may be due to common causes or special causes of variation. Sources of variation in the process (as shown in Figure 15.2) include the operator, the procedure, the measurement reference, the gage or equipment used, the total environment within which the measures are taken, and the interrelationship of all of these variables within a composite system that affects discrimination ability.

Viewing the measurement process within this context, we can approach the analysis of gage capability in a manner much like that employed for process capability. That is, assess the measurement process in terms of the following:

• Its potential to be capable (i.e., a process potential analysis)

• Its short- term capability

• Its long-term capability

Should special causes of variability be identified in these efforts, they should be investigated and eliminated. Common causes of variation, inherent to the measurement process, will provide a basis for capability assessment at each stage of the evaluation process. As with any other production process, the consistency of the measures taken do not necessarily mean that the gage is capable; that is, control is not equivalent to capability at any level of assessment.

There are a number of situations that, if they constantly and consistently occur, may be considered to be "indicators of measurement weakness" (Speitel, 1982) and to signal that deficiencies in the measurement process may exist and therefore should be investigated.

• Frequent occurrences of quality problems are attributed to "unknown" origins. This situation often indicates that measurement capability is not adequate or that the measurement process may be unstable.

• Pronounced dissatisfaction is expressed by the supervisory or operating personnel with their inability to maintain control of the production processes for which they are responsible. This state of affairs is frequently caused by inadequate measurement capability.

• Vague or negative responses are often given to questions regarding measurement capability or about measurement-specification relationship.

In summary, it is essential to note that control and capability in the measurement process are just as important as when in conjunction with any other process; and that as with any other process, these characteristics will not occur or develop in spontaneous fashion. If we are to make sound and appropriate decisions in product and process development, process control and adjustment, or failure analysis, our measurement process must be capable, accurate, and precise. The remainder of this chapter attempts to outline the method by which measurement error may be assessed and gage capability may be evaluated. The intent of this chapter is to outline those procedures required for the analysis of measurement error that is related to the precision of the instrument, rather than identifying and correcting errors that are related to the accuracy of the measurement process. This is due, at least in part, to the fact that calibration procedures (which are intended to correct errors caused by accuracy problems) are relatively well understood , especially in comparison to measurement capability procedures.