How to control when control limit is out of Specs limit

#1
Hello everyone,

Our factory currently manage some characteristics by individual chart. But the special characteristic have Cpk,1.33 so UCL and LCL is out of USL and LSL. Because of technology limit, so we can't handle it completely. I am looking for any suggestions about how to control the process in that case. Note: that all value is in Specs but nearly upper specs, that is the reason why the Cpk is too low.

I am very appriciate with any suggestion from you guys.

Thanks,
Kevin
 
#2
If your control limits are outside the specification limits you'll need to change the process somehow to bring the control limits inside the spec. limits. Otherwise, you'll likely need to do some type of 100% inspection. A lot will depend on what your customer requires. Good luck.
 

Mikey324

Involved In Discussions
#3
Are all values on the high end of the spec, or do they vary from high to low? That would make a difference. As golfman said, a process change to center the values may be all you need.
 

Ninja

Looking for Reality
Trusted
#4
Curiosity, since I don't know what you're making, or how...

In many processes I've been involved in, it is easier to "Mean Shift" than it is to reduce variation....is your process one of these? If your variation is low (like all nearly at upper spec limit) a shift of mean may be all that you need...
 

optomist1

A Sea of Statistics
Trusted
#5
Echoing much of what has been said...and a few more cents:

If the degree of variation is "unacceptable", revise/redesign the process...
In my all too brief experience..with SPC....in general I found it best to follow "Ninja" above i.e. shift the mean re-center...then address any variation issues
 
#6
Hello everyone,

Our factory currently manage some characteristics by individual chart. But the special characteristic have Cpk,1.33 so UCL and LCL is out of USL and LSL. Because of technology limit, so we can't handle it completely. I am looking for any suggestions about how to control the process in that case. Note: that all value is in Specs but nearly upper specs, that is the reason why the Cpk is too low.

I am very appriciate with any suggestion from you guys.

Thanks,
Kevin
How much is the Cp?
 

Bev D

Heretical Statistician
Staff member
Super Moderator
#7
It is most helpful if you can post your data and tell us about the process that creates the characteristic, how it’s measured and what the sampling frequency is...
 

GRP

Involved In Discussions
#8
Our factory currently manage some characteristics by individual chart. But the special characteristic have Cpk,1.33 so UCL and LCL is out of USL and LSL.
A minor observation on this statement: if the control limits are given by 3 sigma and the process is leaning toward USL, then the UCL is not beyond specification. That would only happen for an index < 1. If 1.33 is not high enough for this case, then the UCL is within spec but not far enough from the USL.
 

hngrad1977

Starting to get Involved
#9
If as you say, both the UCL and LCL are outside of the specification you have to much variation in your process and it must be reduced. Based on your brief statement, were there adjustments that were made that would account for the broad range or is this natural machine variation. If it is natural machine variation then it must be reduced or the tolerance opened up. Where I work as a general rule, before we control a process we make sure it is capable.
 

bobdoering

Stop X-bar/R Madness!!
Trusted
#10
You may be in control and not capable. Traditional Shewhart charts don't care about capability - they only care about "signals" of unusual variation to react to. That's why people say be capable first, then keep it capable with control charting.

That said, it is critical to understand your variation. If you are applying normal statistics to a non-normal variation, you will always have wrong conclusions about your process.
 

Top Bottom