An efficient simulation optimization method for the redundancy allocation problem with a chance constraint
针对最小化生产系统成本且系统可靠性作为机会约束的冗余分配问题，提出一种基于信赖域和响应面法的仿真优化方法，数值实验表明其比遗传算法更高效且能找到更优解。
We explore the Redundancy Allocation Problem (RAP) under the objective of minimizing the cost of a production system of general topology in which system reliability is treated as a chance constraint. A novel simulation optimization-based solution method grounded in the concepts of the trust region and response surface methodology is proposed to efficiently solve the generalized RAP (GRAP) under random system survival times. The generalizability of the RAP model and efficiency of the solution method allows for our approach to be utilized in a wide variety of real-world applications. We demonstrate in a series of numerical experiments based on production systems of varying complexity that the finite convergence of the proposed method is much more efficient than the commonly-used genetic algorithm. It is shown that on a simple bridge network, only the proposed algorithm can find the true optimal solution to the GRAP under an allotted computational budget. On a complex network which includes series, parallel, and logical relationships, the proposed algorithm is also shown to find solutions to the GRAP which have substantially lower total system cost than those found by GA under a wide variety of scenarios.