Fuzzy Reliability-Based Large-scale Redundancy Allocation With Penalty Guided Differential Evolution

Abstract

The main goal of Reliability Redundancy Allocation Problem (RRAP) is to find optimal system reliability, ensuring the optimal sub-systems reliability while assigning minimum number of redundant components to the system. Besides, it should also satisfy the constraints such as weight, volume and costs. However, in the dynamic real-world situation, the changes during the optimization process (manufacturing process) lead to uncertainty in the component’s reliability, which can be formulated as fuzzy parameters to ensure trustworthy system reliability. Previous studies addressing this issue considered problems mainly in small-scale RRAP. However, with large systems, the complexity and uncertainty increase exponentially, needing special attention to solve those problems. In this paper, we have considered a large-scale RRAP with twenty-unit system, where the reliability is modelled using fuzzy sets. These large-scale RRAPs are comparatively more difficult than small-scale and they face infeasibility issues during the optimization process. To find the optimal system reliability of proposed 20-unit systems, we have utilized popular differential evolution (DE) algorithm while infeasibility issue is handled using penalty guided approach, collectively its termed as penalty guided differential evolution (PG-DE). The proposed fuzzy reliability-based approach is compared with the traditional model with crisp reliability. The experimental results validate that system reliability produced by PG-DE for the 20-unit system is within a trust region and also comparatively better than the crisp version.