Tolerance limits VS Control Limits
Six Sigma – iSixSigma › Forums › Old Forums › General › Tolerance limits VS Control Limits
 This topic has 6 replies, 7 voices, and was last updated 16 years, 4 months ago by Jonathon Andell.

AuthorPosts

March 13, 2006 at 8:55 pm #42707
Currently using control charts (for avgs and ranges) to help control a manufacturing process.
The sample data are consistently within control limits and the control limits are inside the engineering tolerance limits ie. specifications limits.
I am confused because a high % of product is outside the tolerance limits even though the process is within the control limits?
Can someone provide explanation?0March 13, 2006 at 9:41 pm #134978
Neeraj KatareParticipant@NeerajKatare Include @NeerajKatare in your post and this person will
be notified via email.This is typically happen when your process is in control but not capable to deliver expected performance. Specification limits primarily shows the expectation from process. You have to estimate your process capability and identify the performance gaps.
Other probable reasons could be wrong sample size, biased sample, incorrect sampling procedure, gage error, selection of incorrect control chart.
0March 14, 2006 at 5:27 pm #135037Mark, IF your control limts are inside your specification limits and your control charts are in control and your getting defective product it is because you using an XBar (average) and Range Chart.
Remeber.. The Xbar is plotting the AVERAGE of the supgroup not INDIVIDUAL values. So… I can have one part measure 40, another part measure 0, and my X bar will be 20. If my upper specification is 30 then the part measured at 40 is defective even though the Xbar point is ok.
Control charts are about control, not about capability.0March 23, 2006 at 7:19 am #135400Hi Mark,
I am assuming that you are controlling (thru SPC) some variables to make sure that your product comes out acceptable or meeting whatever tolerance you have. If this is correct, the problem you are seeing right now may be due to the fact that the variables you currently control (via an xbar R chart) are not really significant.
It means that there are other factors that you do not currently control, which are influencing the outcome of whether your product comes out acceptable or not.
Rgds,Ronnie0March 23, 2006 at 11:45 am #135410Excellent point on the difference between control limits based on the sampling mean and the spec limits established for an individual product. One thing that is often over looked when sampling for control charts is the need to take samples based on rational subgrouping. By rational I don’t mean just a temperal alignment (weekly, daily, etc or something like that) but a sampling rational that groups the outputs into samples where the variation “within” the sample is significantly lower that the variation “between” the groups. The effect of averaging large variations into an acceptable output then is reduced.
The other concern I have with the original post is that, often when the Xbar chart is in control but the individuals have a tendency to large swings, this will be shown as an out of control point in the R chart. If you are attempting SPC without looking to the R chart as well… could be the problem.
Finally, if the charts are set up on the output of the process and not on the variable which lead to the output, then the question is one of control in the first place. Monitoring of statistically stable outputs is afterall still a “quality” operation and not control of the process itself, based on monitoring and control of leading process indicators driving that quality output.0March 23, 2006 at 2:26 pm #1354171. Check the sample size, how you get samples (randomization), what the distribution is, etc
2. Check the math
3. Check your equipment and methods of measurement. Do MSA, GR&R. Check the same for inspection/customers
4. Find Cp, Cpk, Pp, Ppk0March 23, 2006 at 7:36 pm #135443
Jonathon AndellParticipant@JonathonAndell Include @JonathonAndell in your post and this person will
be notified via email.“Tolerance” or specification limits, represent the “voice of the customer.” This is what we’d like the process to do.Control limits represent the “voice of the process,” which is how the process currently performs.If you have a statistically stable process with outofspecification outcomes, your process is stable but not necessarily capable. You need to learn how to adjust the mean, reduce variation, or possibly both. Many people use “six Sigma” projects for that.
0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.