Downloaded from https://www.everyspec.com AMCP 706198 Operating Characteristic, even though the terms of s-confidence levels. Such a spec- points look quite different. Once the entire IfIcation, however, must be framed carefully. Operating Characteristic has been generated, The specification ought to depend on the any pair of points which lie on the Operating operational requirements of the system in Characteristic will give the sume plan, even which the components are located. For ex- though the ARL and URL are completely ample, consider the following specifications different. for mean life 0: 3.2.2 CONSUMER AND PRODUCER 1. Compute the 90% lower sconfidence RISKS limit OL.
Since one can be 90% afconfldent that the true mean life is greater than OL, The a and P risks represent the specified If 0L >' 100 hr, accept the lot; otherwime, decision erron assqciated with nominally reject it. good and bad product, respectively. Since lower risks require more testin& a balance 2. Compute the 90% upper a-confidence between the amount of test effort and the limit e,, Since one can be 90% a&-confident cost of a wrong decision is required (Ref. 1).
that the true mean life is leo than 0u, if Ou 4 100 hr, reject the lot- otherwise, In general, for tests in the development accept it, Al stage, the a and P risks may be high, say 20%, and ought to be low for subsequent Specification No. l is equivalent to one in tests, such as production-acceptance tests, which the consumer risk is 10% at a true say 5% to 10%. Thic is so because the early mean-life of 100 hr. Specification No. 2 Is tests usually are scheduled to allow later de- equivalent to ono in which the producer sign changes, and the user at this point is risk is 10% at a true mean-life of 100 hr.
concerned primarily with being assured that The difference between the two tests is the equipment reliability Is not completely apparent. With Specification No. I, only unacceptable. 10% of lots with a mean life of 100 hr will be accepted; while with Specification No. 2, The next tests are those scheduled Im- 90% of the lots with this same mean-life mediately after the first production run. The are accepted. a and P risks for these tests are usually lower than for the development tests, since now For most tests, the niagirtude of a and the equipment design is complete and more P and the number of test observations N are equipments are available for test.
Risks of related, so that specifying any two of the 10% often are specified. quantities determines the third. One ap- proach, for nonsequential tests, Is to specify The third phase of testing Is a sampling a and N and to choose a test which mini- during production to ensure that acceptable mizes P. For acceptance testing, the trend quality is maintained. If the producer Is now is to specify 0 Instead of am If It is continuously meeting the reliability goals, important that both a and P be specified, a limited amount of testing is deKired, and the sample size Is completely determined.
the a and f risks can be higher than for the Because N is discrete, the exactly chosen initial production tests. values of a and Pl are not usually attainable. The a and P are then moved around, for 3.2.3 r-CONFIDENCE LEVELS suitable values of N, until an acceptable set is found. In sequential sampling, a and Test requirements can be specified In are specified In advance, and the sample