Downloaded from https://www.everyspec.com AMCP 7MItN estimate of failure rate is of little concern, * 1/ vr-+ (-.49) Sespecially since lack-of.s-bit is preserved only in linear transformntions. If b is an For small r, the formula is not accurate and s-unbliad estimate of "t a I/", then i/b them is no simple rigorous formula. Eq. 2-69 is a s-biased estimate of A and R exp (-t/I) is well within the limitations of Assumption is a a-blased estimate or survival probability. No, 5. The choice of modification for small Since no one knows which function "ought" r depends on lots of thinp; it is hard to to be s-unbiased, s-bias is ignored. fault any approach with even a modicum of reason to It.
The +1 with the r merely The equation for estimating the fail, re rate keeps the equation from "blowing up" at Ib the maximum likelihood equation r - 0. * r/T (247) In complicated equations, it is usually sufficient to estimate the standard-deviation wheru or coefficient-of-variation of a function. If s-confidenco limits are desired, when usini , estimated failure rate Eq. 2-67, there are no exact ways to get them because (a) there is a lack of knowledge r, total falures about the details of the tests and (b) the method of combining data causes difficulties. T * total test time See par 2-3.2 and 2-33 for details.
If the mission is partitioned Into phases and Example Nos, 20 and 21 illustrate this subphaes, an equivalent failure rate for the statistical model. mission can be developed, if desired, by a simple averaging process! 2.12.3 SYSIEM MODEL -,Vivo a 00, (2-G8) For series systems, the system failure rate Is a linear function of the component failure rates. For linear functions with A-independent where variables, the mean of the function is the function of the moans and the variance of the a w mission phase or subphase function is the function of the variances.
For more complicated systems, see Part Two, _ =- Implies sum over all phases (sub- Desrign for Reliabtifty the paragraph on Para- phases) of a partitioned mission meter Variations Analysis, w fraction of mission time spent In a For further discussion of this and similar models, see Raft. 33-35, but treat any pro- X. a failure rate while In a cedure with skepticism that sems to violate Assumption No. 5 (uncertainty) or pushes The coefficient of variation of ) is (for large Assumption No. I (additive failure rates for P) environments) too far. 2-116 "". - .- .