Downloaded from https://www.everyspec.com AMCP 706-196 For convenience in the ,d6abra, the tests are decision boundary is reached; see Fig. 3-2, discussed in terms of X; the translation to 0 with n replaced by total-test-time, The is easily made. Test specifications are pre- average "clock-time to decision" is shorter sumed to be in the form of a X0 for ARL, if more test stations are used. Table 3-7 and X, for the URL. Specifications given in showb how the average "clock-time to deci- terms of R or 0 can easily be converted to sion" depends on the number of test stations X. Only 2 types of tests are in common use: in the nonreplacement case. I-sample and seq-sample.
The savings in time can be put i,. quantitative The concept of total-test-dine is important. terms, as Table 3-7 shows. To derive Table It is the cumulative operating time of all units 3-7, define E jt ,s , the average waiting time regardless of any censoring or of staggered to observe the lirst r failures from s test- starting times or of replacements. It is to be stations (no replacement), s > r. distinguished from clock (calendar) time, Replacement affects clock-time, but not NE {tr,a} 8 (3-19) total-test-time. The number of test-stations might be limited, or the number of available The entries in Table 3-7 are values of the equipments to go on test might be limited. ratio E It,,/ X tr,..
which quantitatively The test etatistic is always total-test-time, but ra tioe sae, antitaivelydn the relationship among number of equip- of te 4 ments for test, number of test-stations, and clock-time for the test depends on the initial In many cases, the (s - r) units that have number on test, whether test.stations are not failed will still be serviceable. If the keot occupied (replaccment), and othe.r phy- failure distributions of these units are actually si'.d strategies. None of these affects the final decision, ihey affect only the clock-time exponential, the survivors will be as eood as at which the final decision can be mhde.
newt Even id the survivors deteriorate enough to render them unfit for further service, the appreciable savings in time may be worth the 3-7.1 I-SAMPLE cost of the additional units. Virtually all l-%ample plans rre truncated Example Nos. 25 and 26 illustrate the (curtailed), i.e., the test Is stopped when a procedure. TABLE 3.7 RATIO OF s-EXPECTED WAITING TIMES TO OBSERVE FAILURE r IF THERE ARE s TEST-ITATIONS 8 r 1 2 3 4 5 10 1 20 1 1 0.50 0.33 0.25 0.20 0.10 0.30•7 0,050 2 ... 1 0.56 0.3n 0.30 0,14 0.09i 0.008 3 1.... 1 0.59 0.43 0,18 0.12 0,087 4. ......... 1 0.62 0.23 0.14 0,104 5. ........... ... 1 0,28 0.18 0.125 10 ......... ... ... 1 0.35 0.23 C 3-28