On the reliability modeling of weighted k-out-of-n systems with randomly chosen components

Abstract

The weighted k-out-of-n (briefly denoted as weighted k / n) systems are among the most important kind of redundancy structures. We consider a weighted k / n system with dependent components where the system is built up from two classes ${\mathfrak {C}}_X$ and ${\mathfrak {C}}_Y$ of components that are categorized according to their weights and reliability functions. It is assumed that a random number M, ${M=0,1,\dots ,m}$, of the components are chosen from set ${\mathfrak {C}}_X$ whose components are distributed as $F_X$ and the remaining $n-M$ components selected from the set ${\mathfrak {C}}_Y$ whose components have distribution function $F_Y$. We further assume that the structure of dependency of the components can be modeled by a copula function. The reliability of the system, at any time t, is expressed as a mixture of reliability of weighted k / n systems with fixed number of the components of types ${\mathfrak {C}}_X$ and ${\mathfrak {C}}_Y$ in terms of the probability mass function M. Some stochastic orderings are made between two different weighted k / n systems. It is shown that when the random mechanism of the chosen components for two systems are ordered in usual stochastic (st) order then, under some conditions, the lifetimes of the two systems are also ordered in st order. We also compare the lifetimes of two different systems in the sense of stochastic precedence concept. The results are examined by several illustrative examples under different conditions.