Quality of Measurement Data

Measurement data is found to be more useful than ever. There are many ways it can be used, such as making a processing decision that could rely on the results of these data. Another use of this data can also be for creating relationships between variables. 

With this in mind, it is important to note the quality of this data. The quality of this data is directly proportional to its benefits. Higher quality means higher benefits and lower quality means lower benefits. 

The statistical properties often used to classify data quality are bias, variance, and interaction. 

Bias refers to the tendency of an environment. Variance refers to the measure of how expanded the data is. Interaction refers to the relationship of two or more variables which would cause either good or bad consequences. 

Purpose

The primary aim of this document is to provide rules and protocols in assessing the quality of a measurement system.

Terminology

Main Definitions

• Measurement – process of assigning values to physical objects regarding its individual properties
• Gage – refers to an instrument to measure
• Measurement System – the overall process for obtaining measurements 

Minor Definitions

• Accuracy – closeness to the true value
• Bias – sometimes known as a systematic error
• Capability – variability in a short amount of time’s worth of readings
• Consistency – the degree to which repeatability varies over time.
• Effective Resolution – a unit of measure that results in a useful piece of data
• Gage R&R – acronym for Gage Repeatability and Reproducibility, a combination of both R&R
• Linearity – change in bias over time
• Measurement System Capability – estimation of measurement device variance over a short period of time
• Measurement System Performance – the measurement system’s ability to react to changes in the measured function
• Performance – readings taken over a long period of time have a lot of variation
• Precision – repeated readings’ “closeness” to one another
• Reference Value – known as the accepted value
• Repeatability – ability to have successive trials
• Reproducibility – ability to produce the same results
• Stability – a stable measurement process regarding location
• Standard – known as the reference value
• True Value – known as the actual value
• Uncertainty – an approximate number of values across the calculated value that is assumed to contain the true value
• Uniformity – over the standard operating range, the improvement in repeatability

Standard and Traceability 

As per ISO International Vocabulary of Basic and General Terms in Metrology (VIM), traceability is defined as the:

 “The property of a measurement or the value of a metric that allows it to be linked to specific references, typically national or international standards, in an unbroken chain of comparisons all with stated uncertainties.”

With traceability, the need for retest and reductions are greatly reduced.  

True Value

The TARGET acts as the true value of the part. The best scenario for any organization is to have a value that is economically close to the TARGET. However, there are uncertainties which can be minimized by employing a reference value based on a well-defined criteria that can also be traceable. 

The Measurement Process 

It is important to know what’s the primary aim of a process and what are the necessary steps to achieve them. 

Regarding this, the Process Failure Mode Effects Analysis comes into mind. This analysis describes all process-related risks and to come up with methods to avoid negative consequences. The outputs of this analysis are then fed and used for the organization’s control plan. 

A process always begins with inputs, which in this case are the needs of the customers. The customers own the processes and aim to make an excellent decision with minimal efforts.The operation then begins with resources provided by the organization. This mostly comes in forms of new equipment or maintenance of old equipment. It ends with excellent outputs that are traceable and satisfies the customer. Customers may also have the power to control and monitor these outputs. 

Statistical Properties of Measurement Systems

A perfect measurement system is a system that would only produce accurate and exact measurements. However, measurement system results that have desirable statistical properties (zero bias, zero variance, zero misclassification) seldom happen. This truth of life would lead managers to establish less desirable statistical measurement system properties, but these statistical properties define the quality of the measurement system. 

Managers determine statistical properties. These defined properties will serve as the basis for choosing a measurement system. This means that operational definitions and acceptable methods of different statistical properties are required. 

Sources of Variation

One of the principal causes why a measurement system cannot be perfect is because of variation. There are several causes of variation which depend on the situation. There are various methods of presenting and grouping these variations, too. 

The alias SWIPE (Standard, Workpiece, Instrument, Person and Procedure, and Environment) represents the six core elements of a general measuring system. Any factors involving these six must be understood so any risks concerning them can be eased or removed. 

Effect of Measurement System Variability 

The errors in the measurement system (measurement capability) must be minimized and evaluated. The effect of all the variations in the measurement system affects the measurement system performance. 

Effects on Product and Process Decisions

After measuring anything, the next step of action to do is to classify its measurement. The most common groups of measurement results are those in terms of the specification (good) and not (bad).

With this results, the constant goal in a measurement system is to maximize the correct product and process decisions. There are two prime choices to achieve such: 1.) reduce variations by improving the current production process and 2.) reduce measurement errors by improving the current measurement system. 

A proper process control establishes a statistical control, target, and variations. Measurement system variations affect the decision making of an organization, which might lead to incorrect product decisions (mostly due to measurement system errors). 

In mathematical terms, this can be expressed as: 

(Cp)abs2 = (Cp)actual2 + (Cp)msa2

This means that the predicted process capability is equal to the sum of the actual process capability and its variations. To attain a specific process capability, separate factors must be considered. 

New Process Acceptance 

It is normal for an organization to adapt new processes to keep up with the current industry pace. Adapting a new process would mean that a buy-off activity would take place. A buy-off activity is a series of steps that involves research and studies made on the new process’ equipment both at supplier and customer level. 

If the measurement system used during the buy-off activity differs from the organization’s measurement system, this might cause the organization further problems. One of the most common problems in this scenario is when the measurement system used in the supplier level results in a much higher value (gage). This leads to a discrimination in the results.

To minimize the occurrence of such scenarios, the best action would be to have a similar measurement system at both levels. 

Process Setup Control 

Tampering refers to an act of continuously adjusting the measurement system. This action would result in a multitude of variations. A well-known scientist, Dr. Deming first performed the funnel experiment for this method. 

With his experiment results, he concluded the following rules:

1. Unless the process is stable, make no changes. 

2. If changes have to be made, execute changes in equal and opposite directions. 

3. Adjust the process accordingly to reach the target.

4. Adjust the process to the last known point of measurement.