Removing the effect of special causes on Xbar-R chart - Centerless grinding process

[Andrews]

Dear cove members,

The process I am studying is a centerless grinding process. I have an R chart with 7 continuous points on the lower side of the average range (with lesser range during the beginning of the production) after which there seems to be a stable process. Reason for the lesser range during the beginning of the production being the cautiousness of the operator in giving the (manual) feed.

Now to calculate control limits I have to remove all out-of control points and unusual patterns or trends. Should I remove the entire set of 7 points to calculate the control limits?

 

[Jim Wynne]

Dear cove members,

The process I am studying is a centerless grinding process. I have an R chart with 7 continuous points on the lower side of the average range (with lesser range during the beginning of the production) after which there seems to be a stable process. Reason for the lesser range during the beginning of the production being the cautiousness of the operator in giving the (manual) feed.

Now to calculate control limits I have to remove all out-of control points and unusual patterns or trends. Should I remove the entire set of 7 points to calculate the control limits?

If you believe that the special causes have been eliminated, it wouldn't be unreasonable to recalculate, but it would probably be a good idea to maintain a record of what took place so that the knowledge can be used in the future.
I have a question though: if the process is well-centered, relatively small ranges are a good thing. If the ranges suddenly changed because the operator was no longer being cautious, why wouldn't you want the operator to go back to being cautious, and reduce the ranges again? After all, the whole point is to reduce variation.

 

[Statistical Steven]

I do not include the first 7 points in my calculations. I consider it "setup" and not normal operating conditions.

 

[Barbara B]

I do not include the first 7 points in my calculations. I consider it "setup" and not normal operating conditions.

I would remove them, too. The data for establishing a Xbar-R-chart should be collected during the normal process. And then checked for normality

Barbara

 

[cokyat]

I agree with JSW05... u must first eliminate the causes then recalculate the CL's and average...

 

[Andrews]

I agree with JSW05... u must first eliminate the causes then recalculate the CL's and average...

Feed of the wheel is manual. During the beginning of the production run, the operator was running with lesser feed rate resulting in lesser production. He later on achieved the right productivity by increasing the feed rate slightly. Abnormaility was in the first 7 points.

Now from what I could gather from this discussion , I should eliminate the first 7 points from the chart and calculate control limits.

 

[Jim Wynne]

Now from what I could gather from this discussion , I should eliminate the first 7 points from the chart and calculate control limits.

That's the consensus, but keep in mind that you should be as sure as possible that the process is running under "normal" conditions.