Six Sigma: ANOVA Gauge Repeatability & Reproducibility Methodology
ANOVA Gauge Repeatability and Reproducibility (ANOVA Gauge R&R) is a technique of Measurement Systems Analysis. It evaluates the measurement system using the ANOVA (Analysis of Variance) Random Effects model. This evaluation is not only limited to gauges, but also applicable to measurement systems like test methods, measuring instruments, and others.
ANOVA Gauge R&R determines the viability of a measurement system by measuring the amount of variability in the measurements, and comparing it with the total variability. A measurement system may be affected by several factors like:
- Measuring instruments - the gauge or the instrument, and all supports, mounting blocks, load cells, fixtures, etc. Examples of variation sources are sloppiness in mating parts, ‘zero’ blocks, machine’s ease of use, etc. Sources of variation in systems making electrical measurements include analog-to-digital converter resolution and electrical noise.
- Operators - Efficiency of the people to carry out the verbal/written instructions.
- Test methods - include how the devices are set up, parts are fixed, data is recorded, etc.
- Specification - based on which the measurement is being reported. Though engineering tolerance does not affect measurement, it is vital in the evaluation of the measurement system’s viability.
- Parts - what are being measured. While a measurement system may hold good for measuring steel block length, it may not be suitable for measuring rubber pieces.
Gauge R&R consists of the following two important aspects:
- Repeatability - The variation in measurements taken on the same item, under the same conditions, by a single person or instrument
- Reproducibility - The variability induced when different operators (or laboratories) measure the same item
Gauge R&R is used only for the precision aspect of a measurement system. It is an important Six Sigma methodology tool, and is also a PPAP (Production Part Approval Process) documentation requirement. GRR (Gauge R&R) measures parts under the established measurement system, and aims to report all possible variation sources in measurement, for understanding and assessment.
Multiple operators are needed for getting report on reproducibility errors. The ASTM E691 Standard Practice requires at least 10 operators or laboratories. Others demand only 2 or 3 for measuring the same parts. For accounting repeatability errors, one operator measures the same part several times. In case of multiple testing of different parts, full set of operations should be included in each measurement cycle. For accounting operator interaction with different parts, usually five to ten parts are measured. The GRR matrix enables the Quality Engineer to assess risks based on the vitality of the measurement and its cost. There are several methods for determining the degree of replication and sample sizes. The ‘10x2x2’ (ten parts, two operators, two repetitions) is a common sampling for some studies.
The following are some common misconceptions regarding GRR:
- Single GRR is sufficient per family of gauges - GRR relates to a complete measurement system – including the specification, part, operator, and method – rather than one caliper.
- GRR will not pass using parts - GRR can assess precision of a device of system. only if that system is measuring standard weights and blocks. In case of changes in part like deformity occur in measurement, it is taken as a component of measurement system error.
- It is necessary to report GRR results on PPAP documentation for each measurement - It is not necessary, as usually, the Quality Engineer himself makes a good assessment. Only if the characteristic is vital to safety, it requires valid GRR. Otherwise, it is not required for easy measurement of parts with acceptable precision. During PPAP reviews, customers may require additional GRRs.
- GRR performance is very expensive - GRR performance is inexpensive for simple devices, and the result can be utilized for assessment of subsequent measurements using the system. Costs occur for additional measurements only. Only in case of destructive testing, the costs are higher.
- GRR must be within 10% to pass - The final decision to accept the precision of measurement system is taken between the supplier and the customer. GRR is just a tool for assessment, and not a deciding factor whether to accept the level of precision of a measurement system or not.