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 oTh!fAoC 706-19" square distribution with 4 degrees of freedom, The curves for X, would all be the same as required in Eq. 11-5. They are pseudo- (except for scale) if very large samples were random because they exist beforehand (for used. us) on a sheet of paper. Since the numbers are The tests were all terminated at the pseudo-random, a choice must be made onl second failure. Obviously, there is a great deal how to use them. Arbitrarily pick column i of watter in the test results. for X, i = 1, 2, 3; and begin at the top and go This Monte Carlo trial, by hand, has down in sequence.

We will hope that the shown the shape of the central portion (say, method of generating these numbers did not 5% to 95%) of the distributions. Mora trials have a "cycle" such that the rows contain would extend that range. The example was set highly correlated numbers. up to use only one probability distribution Table 11-5 contains the calculations for for the trials; this was for convenience in the A, and for ); X,, is derived from Eq. 11-2. doing hand calculations. In practice the distri- butions need not be the same for all elements. The estimate of X in Table 11-5 occupies the same relative position that the random num- REFERENCES ber does in Table 11-4.

Column 4 in Table 11-5 contains the estimates of the system fail- 1. A. H. Cronshagen, Application of Monte ure rate. At the bottom of each column, there Carlo Techniques in Reliability Ealu- is the sample mean x,_sample standard devia- arlo Tehnqes ineal E, a- tion s, and the ratio six. ation, Aerojet-General Corp., Azusa, Calif., 5 June 1962. As to be expected, the mean of ), is the L. C. W. Churchman, R. L. Ackoff, and E. L. A sum of the means of the X . But the variance Arnoff, Introduction to Operations A Research, John Wiley and Sons, Inc., of ,, is more than the sum of the variances of N.Y., p. 75. the X,. This means that there was some cor- 3.

DA Pam 70-5, Mathematics of Military relation along the rows. A statistical test Action, Operations and Systems. r showed that the ratio 17.82/15.08 = 1.18 of 4. M. L. Shooman, Probabilistic Reliability, the Var {X}/(Var {.}+ Var {'21 4-Var McGraw-Hill Book Company, Inc., N.Y., A 1968. N?' ) would be exceeded by chance about 5. E. Parzen, Modern Probability Theory and 5% of the time; probably not too bad. Its Applicatioas, John Wiley & Sons, Inc., The sample Cdf's ae plotted (smoothed N.Y., 1960. somewhat) in Fig. 11-1, on s-normal distribu- 6. Mathematical Simulation for Reliability tion paper (an s-normal Cdf would appear as a Predictions, RADC Report, Sylvania Elec- straight. line).

Needless to say, none of the tric Products, Waltham, Mass., October distributions are s-normal. The pdf's are all 1961. skewed to the right; there are some very large 7. M. G. Kendall and A. Stuart, The Ad- sample values. The coefficient of variation vanced Theory of Statistics, Vol. IIStatis- (s/i) is more than 1, which also shows the tical Inference and Statistical Relation- skewness of the distribvtions. ships, Hafner, 3rd ed., 1971. ", ЈI-7