|
||||||||||
|
| DataLyzer.com |
|
Gage RR at DataLyzer.com Gage RR by Stephen Computer Services, Inc. Guidelines for acceptance of gage repeatability and reproducibility (%R & R)
* 10% to 30% error - Measurement system may be acceptable depending on application importance, cost of the cage, cost of repairs or improvement and existing process capability. * Over 30% error - Measurement system needs improvement (depending on process capability). Make every effort to identify problems and correct them. Gage Performance Curve Using data from Long Variable and Attribute gage studies, the purpose of a gage Performance Curve is to determine the probability of a measurement system either accepting ore rejecting parts improperly. Once the amount of gage error has been determined via a gage study, it is possible to calculate the probability of accepting a part of some reference value when using the gage. The gage performance curve provides a graphical method of communicating how well the measurement system is discriminating good parts from bad parts as the measurements approach the decision points at the upper and lower specs. Graphical Interpretation The graph contains 3 scales, one along the bottom and 2 along the sides. The one on the bottom contains values as reflected by the measurement system. The ones on the left and right relate to the percent confidence. Specification limits are shown as vertical lines intersecting the horizontal scale at the upper and lower specification values. Vertically oriented diagonal lines are shown intersecting the horizontal scale and the specification lines. These are the confidence indicators. At the intersection of either spec you can read the corresponding left or right scale to see the confidence one has in the measuring system’s result at that size. The measuring system is being used to make decisions about the products being measured. How confident can one be those decisions are accurate? Reading confidence for values Below the lower spec Find the measurement value on the scale at the bottom of the graph. Go directly up to the diagonal confidence line. Then directly left to the percent scale and read the percent confidence of the conclusion the specific reference value being considered is actually below the lower spec. This is the probability that the gage system will lead you to improperly reject a good part (alpha error). Above the lower spec Find the measurement value on the scale at the bottom of the graph. Go directly up to the diagonal confidence line. Then proceed left to the percent scale and read the percent. When this value is within specification, 1 - % scale value represents the probability the measurement system will accept a part that is actually bad (beta error). The upper spec The scale to the right is used. Gage system discrimination is analyzed in similar fashion. The probability out-of-spec values are being judged properly and the chances in-spec values are being judged improperly is also measured by the scale. The values between which the confidence lines intersect the top of the graph are the values between which you can be essentially 100% confident the measurement system is discriminating properly. Short Attribute Gage Study An attribute gage is a pass/fail (go/no-go) type of gage. The short attribute gage study involves two Appraisers taking two measurements on twenty sample parts. Some good and some bad parts should be included in the sample parts. Each appraiser should measure all the parts once then repeat measuring the parts again. If all results for any part do not match, the gage is not acceptable. Conversely, the gage is acceptable if all four trials agree for every sample. If the gage is judged to be unacceptable, it can be re-evaluated using this method, re-evaluated using a more rigorous long form attribute study or it should be repaired or replaced.  Long Attribute Gage Study An attribute gage is a pass/fail (go/no-go) type of gage as apposed to a continuous or variable gage, which renders numeric measurements. Single and double limit attribute gages can be studied using this method but only one limit can be studied at a time. The study consists of one operator making twenty measurements one 8-10 samples. Samples must be selected carefully. The study requires eight to ten pieces, which are measured by a variable gage so their “true” measurement is known. Choose parts at even intervals along the entire range in which gage error might be expected.  Initially, selected samples should range from well inside the limit (such that 20 acceptances are expected to a point well outside the limit such that zero acceptances are expected. Run each part through the gage 20 times and record the number of acceptances. Results must include at least one part with 20 acceptances, one part with zero acceptances and at least 6 parts with between 1 and 19 acceptances (this is often the most difficult condition to meet). Sample selection must continue until all these conditions are met. The probability of acceptance for each of the 8-10 samples are plotted using normal probability scaling with a least squares line drawn through them. No reproducibility analysis is available since there is only one appraiser in the study Gage RR at DataLyzer.com |
| Copyright © 2004 Stephen Computer Services, Inc. All rights reserved. | DataLyzer.com |