Downloaded from https://www.everyspec.com AMcp 706-.17 Exactdy the a~m. computational proedure as in E e I 171 R. Luws and TALL Jaokla, "Optiaization by Direct Search md Syaastic: Raictioe of the Si.t of Serech Resim", AICAE J., was used. The results after Steps I and 3 ase jes in Table 2. Vol. 19. pp. 760-7". Jay 1973. Th total number of functien evaluations was 5362 and the computation time was 7.8 seconds.
APPENDIX DISCUSSION Alorithm for Direct Random Sesich and Inter wlReduction [Equality constraints are presumed to have been eliminated] The negiible computation time for the 15 stage reliability [71 problem shows that the proposed method is very useful for solving reliability problem where disaete units are specified. To emphasize that the recommended procedure does not in- Notation: volve exhaustive enumeration requires only a vety simple calcu- lation. Suppose welookat the posibility ofhavingeither 1,2, x theset ofx, whichare the unknowns 3,4 or S units at each of the 15 stages.
To evaluate all possibili, xeU) the center value of x at iteration/ which cor esponds to ties would require 5 s = 3 X 100 calculations, which is an the best value ofx at iteration j- I. immense, completely impractical number. r the set of r, which are the ranges for direct search at iteration J; the direct search for x, is over the range. ACKNOWLEDGMENTS This work was perforrved with the assistance of a grant from the National Research Council of Canada, A.35 S. Computa- y a pseudo random number, uniform over the range -0.5 tdons were performed with the facilities of the University of to 0.5 Toronto Computer Centre. n total number of iterations, e.g.
n = 100 p number of random trials for each iteration, e.g. p = 100 e the small number by which the range is reduced for each REFERENCES iteration, e.g. e = 0.02 III F.A. Til1man and LM Liituchwaper, "Integer Oramming Formu- iation or Constrained Reliability problems".Maqeemen Scknce, Algorithm: Vol. 13, pp. 81187499. July 1967. 121 ILIL Misa, "A Method ofSolving Redundancy Optimization ?ob. 0. Choose initial values x*() and (f), set / = I. lems",IEEE Trmnt ReL. Vol. R-20, pp. 117-120. August 1971. 1. Calculate p sets xtO) = x,* t J) +y r/I);y is a new pseudo 131 K.B. Mira, "Reliability Optimization of a Seris-Parallel System", IEEE Trens ReL, Vol. R-21, pp. 230-238, November 1972.
random number for each calculation. 141 J. Sharma and K.V. Venkateswaan, "A Direct Method for Maxi- 2. Test thi inequality constraints, retain only those x(l) that mizing the System Reliabiity", IEEE Trens ReL, Vol. R-20, pp. satisfy the sonstraints. Calculate the objective functiop for 256.259. November 1971. each retained x(/). 151 S.K. Banere and K. R jmani "Optimization of System Reliabil- 3. Find the x(l) which maximizes the objective function. Call ity Usig a Parametric Approach", IEEE 7)nt ReL, Vol. R-22. it n*), the center value for next Iteration. If the maximum p p . 3 5 - 3 9 , A p r i l 1 9 7 3 . n u m b e or i t e ra t ion s i s t he s t o p . 161 LB. Misra and J. Sharma.
"A New Geometric Programmin Formu- number of iterations is reached, stop. lation for a Re iability Ptoblem", Int. J. Control. Vol. 18, pp. 497- 4. Calculate r( 1) = (1 - e)r(I). Increment J and go to Step 1. 503, September 1973. 12-24