Abstract: Aiming at the problem that the effective monitoring for project production was difficult to carry out in mold companies due to the large scale of current project groups and the large number of random events,a scheme of mold projects monitoring based on resources decoupling was proposed in consideration of stochastic available resources and activity durations.With classical priority rule,a deterministic project plan with mean activity durations and embedded resource buffers was obtained.According to this plan,the approximate load model of non-monitoring projects was constructed to relieve the constraint relationships between projects.The decision model for monitoring projects was built by using Markov decision processes,which minimized the expected total cost through outsourcing the core activities.Based on reducing the scale of the problem greatly with resource decoupling,the approximate approaches based on duration and resource disruption probability thresholds were further introduced.The computational studies were implemented to verify the effectiveness of the proposed approximate load model,decision model and approximate approaches.

Key words: stochastic environment, project groups monitoring, resources decoupling, Markov decision processes, approximate approach

摘要： 针对模具企业因随机事件众多且在制项目群规模庞大而导致项目生产难以有效监控的问题,考虑可用资源及任务工期的随机性,提出一种基于资源解耦的模具项目群监控方案。利用经典优先规则得到基于任务工期均值且嵌入资源缓存的确定性项目计划,据此构建非监控项目的近似负荷模型以解除项目间的约束关系。结合马尔可夫决策过程理论建立监控项目的决策模型,通过外协核心任务使项目群的总成本期望最小。为应对维数灾问题,在资源解耦极大降低问题规模的基础上,进一步提出基于工期和资源中断概率阈值的近似求解方法。通过计算实例验证了所构建的近似负荷模型、决策模型及近似方法的有效性。

关键词: 随机环境, 项目群监控, 资源解耦, 马尔可夫决策过程, 近似方法