Downloaded from https://www.everyspec.com AMCP 706.198 TABLE 2-3 DATA SET A (Ref. 4) Ordered failure timos (in flight hours) for a pump in the fuel-delivery system of an serial firemsupport heicoptr. N - 20. The entire set of 20 wa failed, so there is no censoring. Columns I and 2 present the original dia, Column 3 shcws the calculatlon for plotting the sample Cdf In the K.S Bounds method. Column 4 shows the 3 plotting pol- tiors for the Beta Bounds method. The remaining columns are Included merely to show how the Ha.wd plotting compares with Cdf plotting in the case where there is no centoring and the two methods can be compared directly. ting pesl. Hazard Cumuls.
Cdf Failure tion, % Reverst Were. tive boe. 1 - sxp (-Hr). Rank timr, hr rAY,% (Talde 21(S)) reok mants H1,2| N 11.3) %M4 _ Cdf plot.
1 175 5 0,26/3.4/14 20 0.050 0.00 4.9 2 605 10 1,8/8.3122 19 0.053 0.103 9.8 3 872 15 4.3/13/29 18 0,056 0.156 14.6 4 1250 20 7.3/18/35 17 0.059 01217 19.5 5 1291 25 11/23/40 16 0.063 0,280 24.4 6 1402 30 14128/46 15 0.067 0.346 29,3 7 1404 35 18/33/51 14 0.071 0,418 34.1 8 1713 40 22/38/50 13 0.077 0,495 39.0 9 1741 45 20143/60 12 0.083 0.578 43.9 10 1893 50 30/48/65 11 0.091 0.6%9 48.8 11 2025 55 352/70 10 0.100 0.769 53.6 12 2115 60 40/57/74 9 0.111 0.880 5865 13 2172 66 44/62/78 8 0.125 1.005 6U.4 14 2418 70 49/67/62 7 0.143 1.148 06.3 15 2583 75 64/72/86 6 0.17 1.314 73.1 16 2725 80 60/77/89 5 0,200 1.514 78.0 17 2844 W5 65/82/92.7 4 0.250 1.764 82.9 18 2890 90 71/87/95.7 3 0,333 2.008 87.7 19 3268 96 78/01.7/98.2 2 0.500 2.508 02.6 20 3538 100 86/98.6/99.74 1 1.000 3.598 97,3 1) Calculetlons rmade to significant fliurte; only 3 decimel place recordod.
21 AN - Vireverse rankl 31 N, - E.AHI 41 This Is the Odf plotting poslitlon that corresponds to the cumulativs hazard; this technique Is used when the Heard scale Is not shovin an the graph~ peper, investigation and/or corrective action shows the results. it to be safe. The long-lived units might be st.'died to find out why they were so good. The same data set will now be plotted by the Beta Bounds method, Cola, 2 and 4 5. If the K-S Bounds method had not from Table 2-3 (Data Set A) are used, see been used, the ancertainty would not have Example No. I (B). been realized.
An engineer could easily have presumed that lipf #5 was the whole story Two more data sets (B in Table 2-4, and and thus misled himself and others about C in Table 2-5) are plotted to help illustrate