hexagon logo
Register Sign in
  • Replies 14 replies
  • Subscribers 223 subscribers
  • Views 149 views
  • Users 0 members are here
Options
Related

Identify Variation

acgarcia
Good Morning,

I don't have a lot of experience is statistics. Can someone guide me on how to identify variation in my measured data? I need to know how much variation is present so engineering can adjust tolerances accordingly and also to make sure the CMM is trustworthy.

I placed a part in a fixture and ran my measurement routine. I did a do/unitl which gave me 6 reports and 6 rows in excel for each part serial number. Data was written directly into excel. Part remained in the fixture for the duration of the inspection.

Here is 1 part with its 6 runs.
































































Serial Run # ID OD TP X Axis Z axis
1 0 3.0768 2.679 0.005 -0.0024 0.0006
1 1 3.0769 2.679 0.0051 -0.0025 0.0005
1 2 3.0768 2.6789 0.0051 -0.0025 0.0006
1 3 3.0769 2.6789 0.005 -0.0024 0.0006
1 4 3.0768 2.6789 0.005 -0.0024 0.0006
1 5 3.0771 2.6789 0.0051 -0.0025 0.0007

Parents
  • Do you have Minitab?

    If you are going to do one part, you could run a Type 1 Gage Study. I think for that study you would run the part 50 times to assess the bias and repeatability of your measurement device.

    P-value would need to be above 0.05 and the Cg and Cgk would need to be above 1.33 to be assessed that your measurement routine and device is repeatable.

    I know had a pretty slick way of doing repeatability tests on his new programs. I think I just saw a thread about that, but not sure which one it was in.

    As for tolerance, determining the Upper Spec Limit and Lower Spec Limit:

    In Excel run a =STDDEV() on your column, then find the average. To determine upper spec limit (for a Cpk of 2) take the Average + STDDEV*6. To determine lower spec limit, take the Average - STDDEV*6.

    If you want a tighter tolerance, then decrease the amount you multiple the STDDEV by. Multiplying the STDDEV by 3 would result in a Cpk of 1.
  • in reply to bfire85
    +/- STDDEV*3 it should be not 6. Given that following Normal destribution +/- 3*STDDEV covers 99.7% of probability.

    In my company we measure 30+ times, take the standard deviation and multiply *6 and divide by the Tolerance range. If the result is lower than 10% we accept the results (repeatability results that is) with no rational. Otherwise we investigate and try to justify the dispersion.
Reply
  • in reply to bfire85
    +/- STDDEV*3 it should be not 6. Given that following Normal destribution +/- 3*STDDEV covers 99.7% of probability.

    In my company we measure 30+ times, take the standard deviation and multiply *6 and divide by the Tolerance range. If the result is lower than 10% we accept the results (repeatability results that is) with no rational. Otherwise we investigate and try to justify the dispersion.
Children
No Data
© 2024 Hexagon AB and/or its subsidiaries. Privacy Policy | Cloud Services Agreement