Hello,

I've been asked to review some probability charts at work (see below chart).

10091

Based on the probability plot, it looks to me that the gage they are using is not accurate enough for this measurement. I've asked them to do a GR&R study on the gage. Are there any other suggestions that I can pass on to this team they can try to get better data? Any suggestions would be appreciated.

Thanks.

Can we have some basic idea of what we are looking at - process, gage, characteristic? Sure would help us not lead you astray... :cool:

The team is looking at the capability of a pneumatic crimp tool (it is like an air-powered set of pliers). They are collecting data from this tool on a gage that measures the deflection and this is coverted to a force. They want to see if the tool can meet capability requirements from a customer based on a ±200 Newton limit.

I've also seen this same thing from our QA people trying to do capability studies on the height of some production part features (stamping process). They are using calipers to do these measurements.

Hopefully, that explains it a little better.

I think you are on the right track. The gage appears to be measuring in 30 Newton increments and may not have adequate resolution.

well a normal probability isn't the correct analysis tool.

30 newton increments may be sufficient given a 400 newton tolerance spread.

You should definitely request some type of Gage R&R study.

*I* use and recommend two readings of 30 parts that span the total variation of the process. calculate the number of distinct categories but more importantly plot the repeated measuremetns on a scatter plot with a 1:1 (45 degree) line and equal y and x axes. (if you have perfect agreement the points will all fall on a 1:1 line. any deviation from the 1:1 line is measurement error σe. this can be estimated by calcualting the average range between the two repeated measuremetns and dividing by d2 (=1.128)

then post your results...for further advice and insight