Reliability Allocation And Optimization Of Multi System Based On Hybrid Topology Adaptive PSO Algorithm

To overcome the disadvantages of premature convergence and falling into local optimum easily of standard PSO(Particle Swarm Optimization) algorithm. Three different styles PSO algorithms are proposed from researching the PSO algorithm in the perspective of mixing algorithm, modifying the velocity update equation and improving population structure. To overcome the difficulty in building reliability model of multi system, T-S fault tree is used to contruct system reliability model, and combining T-S fault tree with those proposed algorithms for the reliability allocation and optimization of multi system.Firstly, in order to overcome the shortages of lower local searching ability of EPSO(Extend PSO) algorithm and the slow convergence speed of uPSO(Micro PSO) algorithm, the hybrid E-μPSO(Extend-Micro PSO) algorithm is gived by mixing EPSO and μPSO algorithm, in early searching period, EPSO algorithm is used to maintain the diversity of particle, when the global optimal solution doesn’t change, μPSO algorithm is used to improve local searching ability.Secondly, the aPSO(New Adaptive PSO) algorithm is proposed to solve the shortages of the existing adaptive PSO algorithm do not consider the attraction and repulsion forces between the particles. At first, to ensure the global searching ability and maintain population diversity, the concept of attraction and repulsion forces based on artificial physics is introduced. Then consider the global optimal solution and the gravitation interaction by other particles with better fitness value to improve local searching ability When the particles entering the global optimal area.Thirdly, in order to further improve the optimal performance of the proposed PSO algorithm, the hybrid topology μPSO, E-μPSO, aPSO algorithm are proposed by mixing the fully connected topology and ring topology. The global optimal searching ability of the three proposed hybrid topology PSO algorithms are proved by comparing to μPSO, E-μPSO, αPSO algorithm by the optimization example.Fourthly, to construct the reliability optimization model of multi system. Taking the multi fault state of complicated system into account, the T-S fault tree is applied to construct the fault probability function, and then constructing either single-objective model with objective function that minimize the system fault probability, which is subjected to constraints including cost, weight and volume, or multi-objective model with more than one objective function such as minimize the system fault probability and cost, restricted by weight and volume. Those proposed algorithms are used to optimize the reliability optimization models.Finally, the method of fuzzy dynamic group decision-making method is gived to deal with multi-objective optimization solution(Pareto optimal solutions set). The fault probability, cost, weight and volume of system are considered comprehensively, the best optimal solution is obtained by the fuzzy dynamic group decision-making method based on fuzzy entropy weight method, topsis method and gray correlation method. The previous optimal solution’influence on the current search scheme is considered for searching the best optimal solution.