# SPC on Non-Normal (Walled) Data

Six Sigma – iSixSigma › Forums › General Forums › Methodology › SPC on Non-Normal (Walled) Data

- This topic has 5 replies, 5 voices, and was last updated 3 months ago by Robert Tipton.

- AuthorPosts
- April 28, 2020 at 11:14 am #247441

scientist doe and validationParticipant@scientistdoeandvalidation**Include @scientistdoeandvalidation in your post and this person will**

be notified via email.Dear all,

I am new to this forum, hope someone can help me out with the question below.

My team is trying to setup statistical process control on a leak tightness measurement. The process generating the data is an inline leak tightness measurement which can give a ‘0’ reading, or a value slightly above ‘0’. For this leak tightness measurement there is an upper spec limit defined, so minor leakage is allowed. Since the data is not normal, but ‘walled’ (with all the ‘0’ values), I think it is not allowed to use the regular rules for statistical process control. Any suggestions on how to still monitor and control this process?

Thanks in advance.

Kind regards,

Ramon

0April 28, 2020 at 11:41 am #247444

Robert ButlerParticipant@rbutler**Include @rbutler in your post and this person will**

be notified via email.I’m not sure what you mean by the regular rules. If you mean the issue of control limits based on standard deviations about the means then you can still use them. The issue is one of expressing the data in a form that will allow you to do this. When you have an absolute lower bound which is 0 the usual practice is to add a tiny increment to the 0 values, log the data, find the mean and standard deviations in log units and then back transform. What you will get will be a plot with asymmetric control limits. Once you have that you can press on as usual.

0May 4, 2020 at 3:03 am #247581

scientist doe and validationParticipant@scientistdoeandvalidation**Include @scientistdoeandvalidation in your post and this person will**

be notified via email.Dear Robert,

Thank you for the reply. I was indeed unsure whether control limits based on the mean and standard deviations would give a meaningful result. Excellent suggestion to log the data, to avoid an LCL below zero. I will try this method on the data and discuss in the team if we could adopt this approach. thanks again.

0May 4, 2020 at 10:58 am #247583

Fausto GalettoParticipant@fausto.galetto**Include @fausto.galetto in your post and this person will**

be notified via email.Do you have any “distribution” of your data, in the interval 0___A?

IF YES, you have to find the SAMPLING Distribution of the MEAN and then derive the LCL and UCL from that.

IF NO, you have

- to ESTIMATE the Distribution of your DATA,
- and from that find the SAMPLING Distribution of the MEAN and then derive the LCL and UCL from that

Which is your case?

0May 4, 2020 at 9:06 pm #247593

Shamshul othmanParticipant@Bagan**Include @Bagan in your post and this person will**

be notified via email.The normality conditions you refer to is a requirement for normal distributions in the parametric statistics. But you can also use other types of suitable distributions or utilize the non parametric statistics or even convert it to a discrete category. However if your Tightness data is an input variable to Leakage as the output variable then using it without normality conditions in control chart is ok.

0May 8, 2020 at 10:47 am #247660

Robert TiptonParticipant@[email protected]**Include @[email protected] in your post and this person will**

be notified via email.I have done hundreds of leak testers, with no problem

assure that you have a maximum of 10% gage resolution.

use normal Gage R & R, with Normal Bell Curves.

Use SPC with Normal or Pearson Curve

0 - AuthorPosts

You must be logged in to reply to this topic.