I have 3 stages of manufacturing. In my first stage Iam practising Xbar -R chart & whereas in my final stage( despatch) Iam checking the outgoing quality by AQL method.
I have learnt that in Rchart if 10% points are going out of control then I can take those ponts out homogenise the data & then go for Xbar chart then calculate Cp & Cpk. If this is true then those points which I have taken out( out of control points) will be physically present in the process & they reapear in the final inspection. Then those components may appear as a part of AQL which I may have to reject.
Please anyone help me how to sortout.
1. Keep using X bar R – I have found they work well on less than stable processes as they smooth out excessive variation, and allow you to react to the real out of control points, not just every blip you would see on an XmR chart. Move to XmR later when the process is more stable – just a suggestion.
2. Only remove the out of control points when you have removed the cause of special variation that it causing them, and you need to use the rest of the data to determine where your revised control limits should be. 10% sounds like a rule of thumb, so be careful.
3. Try this website http://www.odam.osd.mil/qmo/library.htm as they have a good section on control charts (and a big “Thank you” to the US Navy – it’s been handy to use training others!)
Hope this helps.
The R portion of the XBar &R chart determines if a process is out of control or not. Do not remove the out of control points until your process is stabilized.
The R portion highlight within group variation. If this is to large the UCL and LCL will be falsely large therefore ruining the chart.
Perform a RTY(rolled throughput yield) calculation on your process to determine where to focus your improvement efforts.
I agree with James, and I would add the following to the point 2.
“2. Only remove the out of control points when you have removed the cause of special variation that it causing them, and you need to use the rest of the data to determine where your revised control limits should be. 10% sounds like a rule of thumb, so be careful.”
Be aware that, if you removed points that were out of control because you have identified and removed the special cuse(s) of variation, then Cp and Cpk (capability indices) will tell you what can you expect from the process NEXT TIME, when it will run without these special variation, and NOT how the process PERFORMED in this batch. To have a better idea about how your process performed, you should use Pp and Ppk (performance inices) WITHOUT REMOVING the points that are out of control. This estimation is specially valid if the distribution of all individual values is more or less normal shaped (I mean all values from all subgroups, including the values from subgroups that are out of control)
Hope this can help
The forum ‘General’ is closed to new topics and replies.