Downloaded from https://www.everyspec.com AMCQ 706-98 can be calculated, and a sample or that size a preassigned number, e.g., number of fail- drawn. This process is repeated until a ures or number of test hour.. For practical decision boundary is reached. reasons, all life tests are truncated because of economic and scheduling factors. Trun- *Conventional wisdom attributes smallest cated life tests are. especially suitable when * sample sizes to seq-sample tests, largest sam- the failure rate Is constant, If the failure- pit sizes to I-sample tests, and Iin between time distribution is s-normal or lolnornal, to in-sample tests.
However it Is difficult to however, the mathematical difficulties of compare sample sizes because: evaluating the r~sults of truncated tests are quite formidable. This is also true for other 1, Sample size can be a random variablg, faiture distributions with tionconstant failure usually then one deals with ASN. rates that involve more than one unknownI 2. The OC curves for the tests that are being compared ar,э not (and cannot be) Seq-sample tests ought not to be used in ex~actly the same (see pp. 261-262 of' Ref. contractual situations unless the OC curve 23), even though they may be close in some regions of true quality, 3.
The ASN is a function of truc quality, 11oo *It has been shown (see p. 262 of Ref. 23) go 90 that the usual seq-sample test of Wald, has an ASN no larger than any other mample plan So0~ at the 2 true-quality points where the plans --- -Single, N-75, V'1 consumer and producer risk points). If the distribution ofr true-quality or in,-5 coming lots is known rather well, theni con- 4 suit a statistician to optimize the sampling 3 procedure (of course, choosing the criteria "Nfe for optimization 'will be an exciting task in 20 itself). 10 Fig. 3-3 compares the ASN for several o 1 2 3 4 5 a 7 sampling plans.
The u~sual procedure is used True Percent Detective of Submitted for the I-sample and 2-sample plans (i.e,, d Inspection I ots is monitored only at the possible acceptance points). In Fig. 3-3 tho plans are asserted to Notes: Thea p oints an the OC curve. YwhIch am. nominally be roughly equivalent to the I-sample planth r.f llmnre * of N -75, r 1 (1) Producer risk 10%, at 0.7% defective Tabl 3- copare soe l~t'(2) Consumer risk 10%, at 5% detective. of sampling plans. See Met.
I for details of anelysis, 3.5.2 TRUNCATION Figure 3-3, Average A,,mount of Inspect/on A truncated life test is one Iin which testing Unde Sequentia Doubleg Multple is terminated after a random variable reaches anC eur e ntia Sa pln (AY -... ........ -~