Downloaded from https://www.everyspec.com AMCP 70 4I1 Remember that the uctual a, 0 will be appre- RI(T) eiXp (- XiTM) (3-16b) ciably different from those used in Eqs, 3-11 and 3-14. Exact analyseq are "available" for the ARL and URL, rospectlely, for a rli,, some sequential pluns, but they are tedious ability speclficiltion for r, hours. to program ror a computer: Ref, 17 has exact analyses for its plans. but they are tor the The conversion of specified failure rates to exponential distribution., In general, th' probability-of-survival specifications will lead tests computed from Eqs.
3-11 and 3.12 can to exactly the same types of tests discussed be appreciably truncated without actually in pars, 3-6.1 a•d 3-.2, However, because exceeding the nominal a, of the exponential assumption, the &-expected w.iting time boforu a decision is made can be Example No, 24 illustrates the procedure, calculated. For the I-sample case, Table 3.5 can be 34.3 EXPONENTIAL AS8UMPTION used to determine N and c. It Is shown In Ref.
I I that the a-expected waiting time be- The attributes test Is similar to thz bino- rore a decision is reached (as a function of mial cae in terms of test operation atid true s-reliability A) is criteria (Ref, I), ITe majol difference Is that R is replaced by the exponential formula ((3 exp t- X TM) Z 1 {WtIg} -:-WRxRX (3.17) where where TM n mislon time , 1 X failure rate The "amount of testing" can be measured by The term - i any of the following paiameters:
in Ref, I I for many sets of k and N, N a number of litms on test 3.7 EXPONENTIAL PARAMETER, LIFE r w number of failures TESr8 WI w waiting time before a decision The d13tribution of failure.times is expo- (time elapsed from start of test nenttial' to the time a deidson Is reached) T - total number of accumulated test it W - O(.
hours before a decision is reached where If the AR.L and (JUk are specified in terms t - failure time (e.g., time to failure, of failure rate, and If a mission length TM or time between failui es) can be detennined for the item, this type of test might be appropriate, This, in turn, will R) - Sf (t) yield X - a scale parameter (iailure rate), R4(T) n ep (- X•TM) (3-16a) 1/X - 6 - E it)