Downloaded from https://www.everyspec.com -! AMCP 700-1. size is a random variable whose value is not 1.0 predetermined but changes in sucosasive - tests; see par. 3-5, 3-6, and 3-7. I Conventional wisdom gives more impor- tance to the 2 special points (e, ARL) and (A, URL) than is Justified. What Is important is the entire function which relates thue quality to accept probability over the entire - rang of possible qp.al1ty. Par. 3-2.4 explains 0 R, * this function and Its uses. TRUN 8-.RLIASILITY OP LOT F/gure 3.1t. Typ!tvd: Operatlng Charactertitlc 3.2,4 OPERATING CHARACTERISTIC Qurve for ,eIlobbilty Acceptww CURVE aTet, H.: R-R1, HI:
R-R 1 , U Eelfled a and 0 (Rof 1) When two of the three quantities-N, o, and a-are specified, the accept-reject criterion of the acceptance test uniquely Is determined 3-3 PRELIMINARIES TO TESTING for a given family of tests. It is then pos- sible to generate the Operating Characteristic Test designers must consider many factors (OC) curve of the test plan. This curve In addition to statistical ones. Thes engi- shows the probability of lot acceptance over neering factors include test environment, all possible Incoming reliability levels.
Two equipment, and test procedures which must points on the OC curve are already deter- be defined hi advance of the test and care- mined--the a and P voints with their corre- fully controlled In order to ensure valid test uponding reliability levels, ARL and URL, results (Ref. 2). Several areas must be con- respectlv~ly. sidered for a reliability test to succeed In providing adequate data: For example, if the specification is in terms of a survival probability for a given 1. Use of existing information period of time, the general shape of the OC curve is as shown in Fig. 3-1. 2. Selection of test parameters The probability of acceptance is a binomial 3. Test procedures and instructions.
probability parameter, It can be interpreted as the long-run proportion of lots that will be accepted. If, for example, the OC curve shows that a lot with a reliability of 0.80 3-3.1 USING EXISTING INFORMATION will be accepted with a probability of 65%, then in the long run 65% of all lots sub- When designing a test, it is often Impos- mitted with 20% defective Items will be sible to include all possible levels of Impor- accepted, tant factors influencing reliability, For example, if the three environmental factors- Each sampling plan has its own OC curve.
temperature, vibration, and humidity must This entire curve is affected by a change In be considered at four levels each-81 different sampling specifications. combinations would have to be generated for