SPC Data Analysis - Excel Spreadsheet I can use for Diameters and Perpendicularity

[f1guru]

just been reading some threads as i am totally new to doing these studies, i used a chart someone on here had recommended and entered my study of 45 parts the tolerance was 16.000 +0.006 and -0.005 mm the company who did the machining of the components managed to hit 15.999 to 15.997 for all components and the cpk was 26 or something!!!! what am i doing wrong as all dimensions were right thru the top of bell curve and in an ideal world these parts could not be better it also said process was not capable please help!!!
Can someone point me in right direction for an excel spreadsheet i can use for diameters and perpendicularity

 

[Joelbear5]

Re: SPC Data Analysis - Excel Spreadsheet I can use for Diameters and Perpendicularit

A Cpk of 26 doesn't seem reasonable with those numbers. And the data creates the bell curve, not the other way around. If you post your data it can be a pretty simple analysis for those with experience.

If you go up to the top of the forum you'll see a little green bar that says "Post Attachments List". Search for SPC chart and you should find something that will work well.

 

[bobdoering]

Re: SPC Data Analysis - Excel Spreadsheet I can use for Diameters and Perpendicularit

A Cpk of 26 doesn't seem reasonable with those numbers. And the data creates the bell curve, not the other way around. If you post your data it can be a pretty simple analysis for those with experience.

Actually, since the tolerance is 11 microns, the dimension should be manufactured using precision machining. If there is a bell curve with precision machined products, two things are possible;

1) excessive measurement or gage error
2) overcontrol

A precision machined process should have a continous uniform distribution. A normal distribution indicates an out of control process, in that case.

 

[f1guru]

Re: SPC Data Analysis - Excel Spreadsheet I can use for Diameters and Perpendicularit

sorry i got it wrong, the figures i was asked to do it for is surface finish... the upper limit is Ra0.8 and the finish we provided across the 45 samples is between 0.03 and 0.05 so well below the upper tolerance, is this a feasible study or as results are so low the cpk will be this high

 

[f1guru]

Re: SPC Data Analysis - Excel Spreadsheet I can use for Diameters and Perpendicularit

0.0530.0480.0440.0470.0560.0470.0460.0410.0400.0430.0450.0400.0470.0460.0390.0520.0400.0400.0570.0500.0460.0530.0450.0530.0460.0390.0400.0460.0480.0460.0480.0380.0610.0430.0360.0350.0510.0500.0470.036

couldnt get attachment bit to work sorry about how its posted in
there is a comma between each its 0.053, 0.048, 0.0440, 0.047, 0.056, 0.047, 0.046, 0.041, 0.040, 0.043, 0.045, 0.0400, 0.047, 0.046, 0.039, 0.052, 0.040, 0.040, 0.057, 0.050, 0.046, 0.053,0.045, 0.053, 0.046, 0.039, 0.040, 0.046, 0.048, 0.046, 0.048, 0.038, 0.030, 0.061, 0.043, 0.036, 0.035, 0.051, 0.050, 0.047, 0.036,
40 entries

 

[bobdoering]

Re: SPC Data Analysis - Excel Spreadsheet I can use for Diameters and Perpendicularit

Attached is an Excel sheet with the data, as well as the capability analysis from Distribution Analyzer. I also added a capability analysis with the data X10, in case your data entry was erroneous - because the surface finish values were really close to a polish if they were correct.

Surface finish is unilateral, in that the target is zero, not centered. That being the case, a capability index that determines if the distribution is centered - such as Cpk or Ppk - is meaningless and nonsensical. So, Pp and Cp are your best to consider.

The next point is that the value "0" is a physical limit, and the distribution cannot go below zero and be valid. Distributions that behave in this manner that are usually applicable are the beta or Weibull distributions. So, the close you are to "0", the less normal you are. The worse you get, the more normal you get. Whenever evaluating a unilateral tolerance with a physical limitation, curve fitting is a good place to start. This data shows very little difference in the fit between the normal and the best fit distribution. In that case, we can use the normal distribution a s the model. We can also assume that the physical limit is not playing a role in the distribution at this time (you are too far away from "0".)

So, the Pp from the normal distribution would be the valid capability.

 

[f1guru]

Re: SPC Data Analysis - Excel Spreadsheet I can use for Diameters and Perpendicularit

Bob
you are 100% a top guy, if you were here now i would buy you the biggest beer i could find. This forum is full of people who are here to genuinelly help us newbies and not to judge people on inexperience. The world is a learning curve and with the right people helping others along it should make you proud that you can help others!!!

Thanks again Bob for your time and assistance