Downloaded from https://www.everyspec.com -W 706-197 Example. pendent as shown by the calculations in Table A r , A, events that unit UA is failed or 2,3. Eqs. 2-28 and 2-19 are used in the calcu. good h tion. B r , B o events that unit U. is ffalsd or The conditions under which events are good conditionally s-independent are sometimes Let the sample points, events, and associated called common-modes,* and the failures probabilities be as shown in Table 2.1. The which result from severe common-modes are probability of each event, as shown, is the called common-mode failures.
This phenom- sum of the probabilities of each of the sample enon is so important it will be illustrated with points in the event. another example. Are the events A,, B, s-independent? To find out, use Eq. 2-26. Example, Common mode (cause) failure: Notation: pr{AmB}=Pr{(arb)}= 0.158 AF, BF failure events of units UA and U 3 Pr{AF } X PrA{B}= 0.250 X 0.380 =AnB, failure event of 0.0915 the system S. They are not the same (0.158 * 0.095); so T a partitioning of the sam- eevents AF, B, are s-dependent. pie space:
event of a Be- I nign Environment, a High- Suppose there are two possible environ- Temperature Environment, ments, light (event EL ) and severe (event Es), and an Electrical-Transent and that the new sample space, events, and Enviooment. probabilities are as shown in Table 2-2. The event A, and B F are conditionally a-inde- Given: The events A,, BF are conditionally ~s-idependent, given E, (i = B, HT, , ET). Pr(AFIEB = {BFIEB} 6 X 10, 4 ~~~~~~~Pr{FEr 0.9976Is }=1 0 " ,_ _ _ _620 0.380 Pr{AIEm Pr{BFIEHT 1= 1 X 10-1 TABLE 2-1. SAMPLE SPACE FOR EXAMPLE 0F9976 Pr0 4 X 10 4 Cursory inspection of the data shows that U A AG aqb g agbf and U.
are quite reliable if the environment is 10.750 0.528, 0.222 00benign, and that nonbenign environments are ___rare. We first calculate the unconditional fail- ure probability for UA and U. (see Table AF abg b 2-4). It is negligib!y different from the benign 0.250 0.092 0. 8 conditional failure probability. This leads us to believe, reasonably enough, that the effects The number associated with each of the 4 sbale of the nonbenign environments are negligibie. ! po ,itS (,aghq, agbf, afb 9 , *gbf) is the probability of that sample point. But then we calculate the protahilities lhat both UA and Us are failed (see Table The events are defined as AG V JA 2-4).
The situation is now quite different; one AF (afbg 9 .ak) of the nonbenign en vironments is most impor- = F (aebJl, a b) 'Now coiled "comrioom-u 2-6