Artificial neural networks for reliability maximization under budget and weight constraints

摘要

Purpose - The purpose of this paper is to apply a recent kind of neural networks in a reliability optimization problem for a series system with multiple-choice constraints incorporated at each subsystem, to maximize the system reliability subject to the system budget and weight. The problem is formulated as a non-linear binary integer programming problem and characterized as an NP-hard problem. Design/methodology/approach - The design of neural network to solve this problem efficiently is based on a quantized Hopfield network (QHN).It has been found that this network allows one to obtain optimal design solutions very frequently and much more quickly than other Hopfield networks. Research limitations/implications - For systems more complex than series systems considered in this paper, the proposed approach needs to be adapted. The QHN-based solution approach can be applied in many industrial systems where reliability is considered as an important design measure, e.g. in manufacturing systems, telecommunication systems and power systems. Originality/value - The paper develops a new efficient method for reliability optimization. The most interesting characteristic of this method is related to its high-speed computation, since the practical importance lies in the short computation time needed to obtain an optimal or nearly optimal solution for large industrial problems.