Downloaded from https://www.everyspec.com AMCP 70M-11 time before a decision cun be made compared severity-level vs parameter of a distribution to that of a nonreplacement test with the (usuall% the constant failure rate), the un- 4 same number of items on test originally, certainty in extrapolated value ought to be This savings in calendar time Is accomplished determined at the nominal conditions. The at the cost of having to place more items on actual failure rate can easily be uncertain bэ test. If a sequential test is used, it Is usually a factor of 10! See Ref.
24 for the case preferable to plan for a replacement test, where the time-independent failure rate obeys since all items may fail in a nonreplacement the Arrhenius temperature equation, test befor e a decision is made, and more test items will have to be obtained, It is easy to be careless about just what is being accelerated In an accelerated test. This is a practical engineering problem, not Most often, it ought to be a particular param- a statistical one. The required data are the eter in a life distribution. If the distribution actual operating life of each item, regardless has more than one parameter, then the of when it was put on test.
acceleration behavior of all parameterb ought to be justified. For this reason, accelerated 3-5.6 ACCELERATED LIFE TESTS tests are more useful in the delgn/develop- ment stage, There, it Is not the actual life An accelerated test is one in which the that is important, but the failure mode/ test conditions are adjusted to accelerate mechanism itself-ind whether it Is likely to failure, i.e., to be more severe. While accel- occur at usual conditions.
crated tests can be used to discover and evaluate critical weaknesses in the parts or An Important consideration in using an design, their attractiveness in acceptance tests accelerated test for acceptance testing Is that is that the amount of test time is reduced the supplier might design his product to pass since the required number of failures for a your accelerated test, rather than to do well decision will occur relatively early, This in the field. reduction in waiting time is most important for items that have very high reliability goals, Further details on accelcrated tests are because the amount of test time required to given in Chapter 7 "Accelerated Tests".
establish conformance can be very large, 3.6 BINOMIAL PARAMETER If the stress conditions are arcelerated, the reliability goal bader standard stress condi- This paragraph describes the statistical tions had to be modified accordingly, l'here- characteristics of tests to which binomial tore, the relationship (approximate) of rell- distribution theory applies, For reliability, ability to the acceleration factor must be the test characteristic of interest is often the known, so that appropriate test criteria can fraction-defective (failed) over some fixed be established.
Most accelerated life tests time of operation, are performed at the part level because of the strir•i-nt reliability requirements existing If Ro represents the ARL and 91 the URL, at this level and because stress/failure relation- the test specification is of the following form: ships are relatively, easy to determine through experimentation, MO.: NO (3-6a) It pays to repeat the warningsj about extrapolation from experimental data. If a regression line has been determined for HI: R 1 (3-6b) A A _,' i 1 I i ..