An improved particle swarm optimization algorithm for the reliability-redundancy allocation problem with global reliability

Most of the existing works on reliability???redundancy allocation problem (RRAP) focus on source-to-sink systems. This work formulates the RRAP with global reliability (g-reliability) called the g-reliability RRAP (gRRAP), which considers imperfect nodes and regards nodes and links of the system as subsystems. The g-reliability is the probability that all nodes of the system are connected through normal links, which is more suitable for the complex modern network systems featured by multi-sources, multi-sinks. We propose the algorithm to calculate the objective function of the gRRAP. And to solve the gRRAP, an improved particle swarm optimization algorithm with a specific particles initialization approach (called PIPSO) is developed. In the specific particles initialization approach, we propose a metric to evaluate the distance between two particles under the RRAP. The experimental results show that the PIPSO algorithm has better performance than several popular meta heuristic algorithms and previous works in solving the gRRAP and four benchmark problems. Most importantly, we propose a measure to evaluate the uniformity of particle position when particle swarm optimization is applied to RRAP. Results demonstrate that, the particle swarm optimization with more uniform initial particle distribution converges faster and better, and the obtained solutions fluctuate less.

Automated summary

This study proposes an algorithm for the redundancy allocation problem (RRAP) with global reliability, which takes into account the characteristics of complex modern network systems. An improved particle swarm optimization algorithm is developed and shown to outperform other meta-heuristic algorithms in solving the problem.