Caster
14th December 2004, 12:11 PM
I have inherited a CSP-1 plan. We x-ray castings using this plan. It was set up before my time by someone who is no longer with the company. I now have to try to justify its use to a customer.
I have spent a lot of time on this site as well as the excellent site by Stan Hilliard www.samplingplans.com (http://www.samplingplans.com/).
The plan seems unique in that there is no “lot submitted for inspection” as such. Instead, a sample is taken from a moving conveyor, and an acceptance decision is made based on that sample. If there are defects in the sample, you go to 100% inspection until “i” or the clearance quantity of parts checks OK. This allows a return to sampling from 100% inspection. Can anyone explain the statistical basis for this?
This design is well suited to the way parts move through our process, and I would like to keep it.
However, it seems to me that the way we use this plan is overly “producer friendly”. It also appears to me to have a very high consumer risk. I want to determine the consumer risk for this plan. And that is a problem for me.
The copy of the standard I have is almost illegible. I can’t easily read the OC curves.
I would like to develop my own OC curve, but I don’t know what distribution this plan is based on.
I’m confused by the use of a sampling frequency instead of sample size
It seems that I first select a sampling frequency code letter based on the number of units in the production interval (Table I). For example for 501-1200 units I can select code letters A through F.
This then lets me select sampling frequency “f” from ½ all the way to 1/10 (Table II-A).
For example if we had 750 units in the production interval at a frequency of 1/10 my sample size is 75? Can I then make an OC curve using this sample size and Poisson method?
Or does anyone know how to generate the OC curve for a CSP-1 plan?
Failing this, I may default to good old MIL STD 105. Defining a lot and a sample from that lot is going to be a challenge however. Any ideas for this when parts flow by on a conveyor?
Thanks as always!
I have spent a lot of time on this site as well as the excellent site by Stan Hilliard www.samplingplans.com (http://www.samplingplans.com/).
The plan seems unique in that there is no “lot submitted for inspection” as such. Instead, a sample is taken from a moving conveyor, and an acceptance decision is made based on that sample. If there are defects in the sample, you go to 100% inspection until “i” or the clearance quantity of parts checks OK. This allows a return to sampling from 100% inspection. Can anyone explain the statistical basis for this?
This design is well suited to the way parts move through our process, and I would like to keep it.
However, it seems to me that the way we use this plan is overly “producer friendly”. It also appears to me to have a very high consumer risk. I want to determine the consumer risk for this plan. And that is a problem for me.
The copy of the standard I have is almost illegible. I can’t easily read the OC curves.
I would like to develop my own OC curve, but I don’t know what distribution this plan is based on.
I’m confused by the use of a sampling frequency instead of sample size
It seems that I first select a sampling frequency code letter based on the number of units in the production interval (Table I). For example for 501-1200 units I can select code letters A through F.
This then lets me select sampling frequency “f” from ½ all the way to 1/10 (Table II-A).
For example if we had 750 units in the production interval at a frequency of 1/10 my sample size is 75? Can I then make an OC curve using this sample size and Poisson method?
Or does anyone know how to generate the OC curve for a CSP-1 plan?
Failing this, I may default to good old MIL STD 105. Defining a lot and a sample from that lot is going to be a challenge however. Any ideas for this when parts flow by on a conveyor?
Thanks as always!
