When or how often should we update SPC (UCL and LCL) Control Limits? Every month?

[alfawei]

In SPC

Such as Xbar-R Chart

The process is stability

The control limit (UCL LCL) need be updated everymonth or every 2 months?

 

[e006823]

Wheeler suggested some rules for when to change control charts. I’ve paraphrased them below:

1) Process exhibits obvious change over past behavior
2) You know what has caused the change
3) You like the change (process has improved)
4) You have reason to expect the change to last

Here is a link to the article:

https://www.qualitydigest.com/may/spctool.html

Regards,
Bob

 

[Darius]

alfawei In SPC

The process is stability

I think you mean stable, on a stable process there is no change of the mean and the variation, so no need to change control limits.

I know Wheeler proposal but IMO control limits must represent the process behabiur, if a pump is changed and the variation of the new pump is greater (worse), the control limits must represent the change of the process altho the change was for worse, I know it's not a traditional way but I change the process when a change happen. In our company we compare the mean and standard deviation between the actual (ready to be closed) and the last closed period with a "t" and "F" tests, if a change is detected a no stability flag is displayed and the control limits represent the new closed period. How many data point is a period? a minimum sample size is chosen.

 

[Rob Nix]

Simply put, you change the limits when the process (reflected in the charts themselves) indicates you should. So only point #1 in e006823's post is applicable (the other points will show in point #1).

There is NEVER a set time period (e.g. week, month) where limits must be recalculated.

 

[e006823]

Simply put, you change the limits when the process (reflected in the charts themselves) indicates you should. So only point #1 in e006823's post is applicable (the other points will show in point #1).

There is NEVER a set time period (e.g. week, month) where limits must be recalculated.

As Rob stated the chart itself will tell you that something has occurred with your process. You still have to determine if the change is “good” (has the process improved), is the change permanent or was it caused by a transient event, and what caused the change.

1) Process exhibits obvious change over past behavior
If this is not true then there is no reason to recalculate the limits

2) You know what has caused the change
If the answer to this is no you should be looking for the assignable cause.

3) You like the change (process has improved)
If this is not true you should be looking for the assignable cause.

4) You have reason to expect the change to last
If this is not true you should be looking for the assignable cause.


It may be a philosophical difference but I believe that control charts should be used to understand and improve your process and unless you can answer all of these questions you have no valid reason for recalculating the control limits.


Regards,
Bob

 

[alan smith]

How often should we update SPC control limits?

As Rob has stated you only recalculate control limits when the process has changed.
Here are some of the guidelines I put in place whilst using SPC to control the process.
Firstly, you need to consider is the change in the process data range good or bad.
If the changes are good and stable i.e. no trending in the data then you can recalculate the limits but you will also need to understand what has changed in the process.
Also consider the process activators. People, equipment, Environment, Materials & methods. If the process has become better and more consistent, then 1 of the activators listed above has changed.
If however the process has got worse then you should recalculate the limits based on the data you have but try to understand what has changed to make the process variable. Again consider the activators.
Ideally, if the process is stable use the control limits but also place warning limits on the Xbar & Range data chart. These limits will indicate a warning point in the chart but do not react to a warning point until you have 5 successions of warning data.
In other words there are no set guidelines when limits must be recalculated.

Hope this helps
Alan

 

[Bill Pflanz]

Grant & Leavenworth talk about revision of Central lines and Control limits in Chapter 4 of their book Statistical Quality Control. They had the following comments:

1. If the process has been in control with both average and dispersion satisfactory for the spec limits then the initial (trial) control limits can be extended to apply to future production.

2. As more data accumulates, the limits are reviewed and revised whenever necessary. "It is desirable to establish regular periods for this review, such as once every week, once every 25, 50 or 100 subgroups." (My opinion is that the length of time for review is dependent on how often there is a significant change in the limits and is a judgement call.)

3. Grant & Leavenworth considered out of control problems in the R chart to be more significant then on the X-bar chart. If the R chart is out of control then they recommended estimating the value of R bar and the standard deviation that might be attained if the dispersion were brought back into control. They also discuss what to do if both R and X bar are out of control.

If more information is needed on the revised limits, they provide more description than I have here plus they have an examples of the methods for revising the control limits.

Bill Pflanz

 

[Charmed]

SQC by Grant & Leavenworth

Grant & Leavenworth talk about revision of Central lines and Control limits in Chapter 4 of their book Statistical Quality Control.
Bill Pflanz

Dear Bill:

Nice to see reference to Grant and Leavenworth. That's my bible for SQC.

Charmed