Xbar-R Chart - Need help with analysis

[PeterWang]

Hi Friends,
Could you give me your comments of attached X-bar-R chart? I need do root cause analysis.The part specification is 60+/-0.5.Thank you very much.

[Stijloor]

In Reply to Parent Post by PeterWang

Hi Friends,
Could you give me your comments of attached X-bar-R chart? I need do root cause analysis.The part specification is 60+/-0.5.Thank you very much.

It will help our experts if you can describe the process. Can you be more specific?

Stijloor.

[adamsjm]

I would start by looking at your data collection process.
There is very little variation in the data per data series, except if it is near a high valued data series. There are no values less than 59.83 (which should naturally occur) and there are too many 59.83 values. The high data values do not look correct. The control chart suggests significant X-bar shifts, but the underlying data is close to a normal distribution.
Were the parts measured in series or after they lost their time sequence?

[bobdoering]

In Reply to Parent Post by Stijloor

It will help our experts if you can describe the process. Can you be more specific?

Absolutely! We need to know the process and if any adjustments were made during process. We also need to know the type of characteristic - is it a diameter, length, or other? We also nee to know the measurement technique - as it can often provide more variation than the process itself! By looking at the control limits, this looks likes like a classic sample not representing the total process variation problem. If the evidence shows that to be accurate, then either the charting methodology or the sampling size needs to change. If this is truly the case, an X MR chart attached shows that you have little to apply a corrective action to - except to correct your charting methodology. Even so, without knowing your process and contemplating longer term variation, this may not be the "final answer." We must never blindly assume X-Bar/R is applicable to every situation. It isn't... for some processes it is a terrible option.

[Bev D]

good feedback so far. I would add to adamsjm's comments that you have low within subgroup variation compared to your subgroup to subgroup variation. this looks like you do not have a rational subgrouping scheme. please tell us teh process and how you are sampling it. This is the only way we can be of more help.