ARTICLES
Gauge Stability
BIS.Net Team
The objective of a Gauge Stability Study is to see how consistent repeatability, reproducibility and bias is over time. This may be achieved in several ways.
Classical
Obtain a reference value from a sample that is representative of the production parts. If none is available, then select a part that falls around the midpoint of the production parts and call it the master sample.
On a periodic basis, daily, weekly, or monthly take a convenient number of readings, such as 5 and plot the information on a standard X-Bar and Range or Sd chart. If there are no out-of-control signals the process is deemed to be stable. Out-of-control points on the x-bar charts indicate instability with bias. Out-of-control points on the Range/Sd charts indicate instability with repeatability
Used to study bias stability
Used to study repeatability stability
This analysis should be done with the BISNet SPC app, which is focused on control chart technology and includes distribution optimized machine power technology in case data is non-normal, affecting especially the Range and Sd charts.
If multiple parts are used then a control chart needs to be applied to each part.
Machine Powered
An alternative stability analysis can be performed with Dynamic Plateau and Dynamic Variation charts which are found in the BISNet Change Analysis app. These are not only more sensitive but provide better information that can better identify causes for the instability. These charts show the onset, duration and magnitude of changes in bias and reproducibility.
Used to study bias stability. More informative
Used to study repeatability stability.
Another option is Dynamic Process Performance charts which show changes in bias, and reproducibility on the same charts, such as shown below. These are also found in the BISNet Change Analysis app.
The above shows changes in bias and repeatability for the same operator.
The best option is to use the Fingerprint Analysis stacking Dynamic Plateau and Dynamic Variation charts or Dynamic Process Performance charts for different operators on top of each other. Ideally every appraiser should have similar patterns and magnitudes. If not, then reproducibility is unstable.
Fingerprint analysis, such as shown in the above image, can compare stability in bias and repeatability for different appraisers. Differences in patterns will show reproducibility instability. Normal significance testing may not show these instabilities.
Reproducibility stability can also be effectively monitored with machine powered stream charts, where each stream is a different appraiser. This analysis is performed using the BISNet Special Analysis App. The same app can be used to obtain a stability index, using Machine Power. There is no such Index available using Classical Technology.