US Army Materiel Command AMCP 706-197 — AMCP 706-197 Reliability Prediction Handbook — US Army Engineering Design Guide PDF

Downloaded from https://www.everyspec.com AMCP 706-197 J The first term in Eq. 2-20 is an upper N events are s-independent if and only if bound; adding terms in succession provides an for every intersection oi events taken 2, 3, ... , alternating series of bounds which get increas- N at a time, the probability of the intersec- ingly better, until exactness is reached when tion of those events is the product of the all terms are used. probabilities of the individual events. This can be a complicated concept; see the Bibliogra- Pr{AnB}= Pr{AIB}Pr{B}= PrBIA}Pr{A} phy at the end of Chapter 1 for a further (2-21) discussion. Eq. 2-21 is a form of Bayes' Theorem. Example.

* Suppose there are 2 units (from one popu- C {{lation) in a subsystem and both must fail for Pr{A.nBnC } = Pr {AJ(BrC), Pr { BIC1Pr {C ) (2-22) the subsystem to fail. If the probability of N i failure of each is 0.200 and the probability of N A subsystem failure is 0.200 X 0.200 = 0.0400, Pr{A = Pr{AI (223a) then the failure events are s-independent. N9 , Even if the probability of subsystem failure N( were 0.0404 (e.g., 1% above the 0.0400 fig- PT{A}= Pr{AIB,}Pr{B} (2-23b) ure), the failure events could be considered J- I s-indenendent for engineering purposes.

Pr{AB - Pr{BIA, }Pr{A, I Suppose that the probability of failure of NB (2.24) each unit is 1.00 X I0' and th probability , Pr{BIAj }Pr[Aj} of subsystem failure is 1.00 X 10"; then the .,= B Afailure events are s-independent But if the probability of subsystem failure were Eq. 2-24 is a form of Bayes' Theorem. 0.000401 (0.0004 more, just as in the pre- coding paragraph), the failure events wo,'d in 2-3 s-INDEPENDENCE no way be s-independent. When failure proba- bilities are very small, one must be very care- Tnere are several equivalent definitions of ful not to ignore events whose probabilities s-independence. From an engineering point of might ordinarily be neglected.

view, the most satisfactory definition is Eq. 2-25. 2-4 CONDITIONAL s-INDEPENDENCE A and B are s-independent if and only if Pr{AIB}= Pr{AIB}= Pr{4 }. (2-25) A very important concept is conditional That is, the probability of A is the same re- s-independence; i.e., two (or more) events can gardless of whether we know that B has be conditionally s-independent, given a par- occurred, or has not occurred, or we do not ticula event. All general theorems on nroba- know about B - B just doesn't make any dif- bilities are valid also for conditional probabil- ities about Bepc -- Bn justcla doesnt mae ndf ference. There are several equations that are ities with respect to any particular event C,.

logically equivalent to Eq. 2-25, each imply- Thus Eq. 2-25 becomes ing the others. (The -- cond equation in Eq. Pr{AI(BnC 1 )} = Pr{AI(BnCi) (2-27) 2-25 actually is implied by the first one.) The = PrIAIC, most satisfactory definition from a statistical and Eq. 2-26 becomes point of view is Eq., 2-26. A and B are s-independent if and only if P{fBI 1 =rAC}P{II.(-8 In many engineering situations, if two events Pr{AnB}= Pr{A l Pr{B}. (2-26) A and B (say, failures) are not s-independent, Eq. 2-26 is defined even for Pr{ B I = 0 or 1 they will be conditionally s-independent, whereas Eq. 2-25 is not.

The extension to given each event of a set of events which is a more than two events is easier with Eq. 2-26. partitioning of the sample space. 2-5