Existing research about the maintenance optimisation of production systems with intermediate buffers largely assumed a series system structure. However, practical production systems often contain subsystems of ring structures, for example, rework and feedforward. The maintenance optimisation of these complex systems is difficult due to the complicated structure of maintenance policies and the large search space for optimisation. This paper proves the control limit property of the optimal condition-based maintenance policy. Based on the control limit property, approximate policy structures that incur a smaller policy space are proposed. Because the state space of a production system is often large, the objective function of the maintenance optimisation cannot be evaluated analytically. Consequently, a stochastic branch and bound (SB&B) algorithm embedding a sequential simulation procedure is proposed to determine a cost-efficient condition-based maintenance policy. Numerical studies show that the proposed maintenance policy structures can deliver a cost-efficient maintenance policy, and the performance of the SB&B algorithm is enhanced by the inclusion of a sequential simulation procedure.

关于具有中间缓冲器的生产系统的维护优化的现有研究在很大程度上假设了串联系统的结构。但是，实际的生产系统通常包含环形结构的子系统，例如返工和前馈。由于维护策略的结构复杂且优化的搜索空间很大，因此很难对这些复杂系统进行维护优化。证明了基于状态的最优维护策略的控制极限性质。基于控制极限属性，提出了导致较小策略空间的近似策略结构。由于生产系统的状态空间通常很大，因此无法通过分析评估维护优化的目标功能。因此，随机分支和边界（SB＆B）提出了一种嵌入顺序仿真程序的算法，以确定一种经济高效的基于条件的维护策略。数值研究表明，所提出的维护策略结构可以提供一种具有成本效益的维护策略，并且通过包含顺序仿真过程，可以提高SB＆B算法的性能。