[ | Next | Previous | Up ]

Re: Reduced sampling plan for a continuous process

From: Stan Hilliard
Date: 21 May 2002
Time: 23:15:31

Comments

Greetings Carl,

I think of the continuous process as one where there can be natural lots but you won't discover their boundaries until after inspection.

SAMPLING THE CONTINUOUS PROCESS: One possible strategy would be to adapt a poisson sequential sampling plan to the continuous process defect count as follows:

Start a poisson sequential plan at some jumbo (number 1). The normalized count (defects/100 sq. yds.) from this and each successive jumbo (number 2,3,4,...) will add to the accumulated count of the poisson sequential plan. When the decision rule makes an accept/reject decision start the plan over, with the next jumbo as jumbo number 1. The number of jumbo's per decision will vary as described by the Average Sample Number (ASN) curve of the specific plan.

The advantage of this scheme is that you can make accept/reject decisions without defining the lots in advance.

The only time that you need to define a lot is after a rejection -- for the purpose of dispositioning the material. The task of defining the beginning and end of the rejected "lot" can be judged from the Cusum of the normalized jumbo counts.

You can use the software program TP105 to design the sequential sampling plan for counts:

www.samplingplans.com/software_oc.htm

Poisson sequential sampling plan examples, including decision rule tables, diagrams, OC Curves and ASN Curves:

www.samplingplans.com/outputpoisson.htm

Tutorial on sequential sampling in general:

www.samplingplans.com/modern3.htm

One down side of the continuous sequential approach is that some jumbos will be in the continue-zone until subsequent jumbos cause the sequential plan to accept or reject. Consequently, some manufactured product will not be dispositioned immediately. That is the nature of continuous plans.

SKIPPING JUMBOS: In addition to using the poisson sequential plan for the continuous process, you could implement it with a skip-lot strategy. This would involve considering each jumbo to be a "unit" in a CSP-1 plan. When run in combination, the sequential plan will make the acceptance decision and the CSP-1 plan will tell you which jumbos to test.

Stan Hilliard


Last changed: November 20, 2007