00342 de71e089a262a6029d4137528ec356

346

Prairie & Zimmer

with the reliability requirement. This check is madę to allow for continuation of sampling if the cumulative number of failures is consistent with the reliability reąuirement.

The plans presented in this paper differ from those of Dodge (1943, 1947, 1951) in that their major purpose is to monitor product on a sampling basis so that if a departure from the reliability requirements occurs, it will be promptly detected so that appropriate action can be taken. The primary method of characterization of the present plans is the average number of defective units accepted for shipment for diiferent sampling rates and unit test times as a function of decrements in reliability from the requirement. This method of characterization is viable for both destructive and non-destructive testing. Also, the present plans allow for continued sampling on a probational basis when the obseryed failure ratę, based on available data, is in conformance with the reliability reąuirement.

A selection of values for the parameters of the plans and a rule for determining concordance are presented. The parameters of the plans are related to the AQL and LTPD concepts of acceptance sampling and the average number of defective units passed before a defective unit is sampled and detected.

Introduction

The purpose of this paper is to describe a procedurę that can be used to design monitoring plans for product from continuous production. The primary objectives of these plans are to monitor the ąuality and reliability of the product for the early detection of defectiveness that is in conflict with the reliability reąuirement of the product. A secondary objective of the plan is to control the outgoing ąuality of the product by inspection for situations where the testing is non-destructive.

In our experience we have found that "błock" sampling plans are often employed for continuous production by manufacturing facilities to monitor the ąuality and reliability of their product. Błock sampling is done by arbitrarily grouping a fixed number of units and sampling one or possibly two units from that błock. The destiny of the błock depends upon the test results of the units sampled. This type of sampling is not consistent with continuous production and there can be a large backup of units if the production ratę is Iow, if the blocks are relatively large, if the test time is long, or if the sampled unit is at the end of a błock.

It is assumed that the units available for acceptance testing are completed units ready for shipment to the customer except for a finał inspection