The research on multi-objective optimization method of system reliability based on the genetic algorithms

In this paper, we consider a series-parallel system to solve the optimization problem of reliability redundancy with two different objective functions, the entropy and the reliability, by analyzing the advantages and disadvantages of the current methods used to solve optimization of system reliability redundancy. We present the algorithms and processes to settle the multi-objective optimization problem of reliability redundancy based on the Genetic Algorithms. The method includes the following two advantages: (1) We solve the multi-objective optimization problem by assigning a weight to each of the objective function then integrate them so that the problem is converted to a single objective function problem; (2) Based on the Genetic Algorithms, we choose the weights randomly. In general, the different weights can result in different solutions, even a very small perturbation in the weights can sometimes lead to quite different solutions. At last, we conduct the simulation by using a typical 4-stage series-parallel system. It is concluded from the simulation results that the GA used in this paper can get higher values of reliability and entropy, meanwhile the optimal solution doesn't vary in spite of the various choices of weight of each objective function. Compared with the existing work in which they use the Global Criterion Method, which has the difficulty in choosing weights, the method used in this paper is better.