Downloaded from https://www.everyspec.com AWCP 70-1MI very loosely, Usually (and improperly) the same; if•> 1, the reliability is getfing li.verse of failure rate is meant. wont.) This is equivalet to saying that log failure rate is a linear function of log time. 2. It is not clear whether the failure rate of the item is constant during the test Occaslonally, thii model is referred to as (Poluon process) or whether the failure rate "Welbull" because of the similarity of this changes during the test (nonhomogeneous nonhomogeneous Poisson process failure- Poisson process-see Ref. 3, pp. 12S-126). occurrence rate to the Weibull failure rate; but the basic models for the two processes 3.
The rate at which resourcem are de- are quite different, voted to the project must stay reasonably constant, as a function of test-time, This Notation follows: requirement eliminates major redesign as well as m•lor shifts In effort. t = time 4, Different (and possibly incorrect) T time at which a prediction Is statements of the exact situation reflect the made (can be t,,) ambiguity In how time is accumulated and how the nonhomogeneous Poisson process time at which falluto I occurred, is precisely defined. a random variable Only one of the possible Duane models is A number of failures in the Interval an&lyzed; another is merely referenced.
The (0, TI, a random variable unless bL•lc assumption of constant effort (resource T - rate) is probably the limiting factor in a satisractory prediction; it is difficult to dem- M(t) mean number of failures In the onstrate, and a poor-fit or the data to the interval (0, t) model is sometimes used au evidence that the etTort is not constant, The arithmetic in the ?10 a faIlure-occurrence rate (intensity first model is somewhat easier; so perhaps it of Poisson process), X(t) N deserves to be tried first, dm(t)/dt; sometimes called "peril rute" 10.2.1 DUANE MODEL NO 1 - constant in the equation for The following assumptions apply:
M(W) 1, The equipment failures are described u(t) ia measure of the uncertainty In by :i nonhomogeneous Polsson process (the 9n X(t) failure rate Is a prescribed function of time). implies an estimate or prediction, 2. 'rime is the cumulative test-time of all See Appendix B for details units on test; e.g., if there are 3 unite on test, time cumulates at 3 times the clock- L, U subscripts which imply upper calendar rate. and lower uncertainty limits on At 3. The cumulative number of failures is directly proportiondl to (Time)Y, where 0 is From Appendix B, which contains the a positive constant.
(If 0 1, the reliability mathematical details, the solution for the is growing; if 1 • I, the reliability stays the parameters of the model is: I.Who