Ranking of elements in system reliability modeling: The least influence method

In this paper, we propose a new method of elements ranking in system reliability design using the theory of fuzzy relations. The problem is formulated as an automatic classification based on the transitive closure of the fuzzy similarity relations. This allows splitting the set of system elements into disjoint classes, which are similar in importance. To define the fuzzy relations of similarity we present the parameters of each element as a vector of influences. To measure the similarity of elements, we use the distance between the two vectors and for the degree of influence evaluation, we present a special algorithm using the expert knowledge about the least impact of an element and about the comparisons to the other influences using the 9-point Saaty scale. The proposed method relaxes the assumption about the independence and the binary-state (up/down) of elements. The possible fields of application are systems with ill-defined structures and multifunctional elements such as organizational, ergatic, military, political and other complex systems.