【作者】 田文迪；

【导师】 崔南方； Erik Demeulemeester；

【作者基本信息】 华中科技大学 ， 管理科学与工程， 2011， 博士

【摘要】 随着经济全球化和市场竞争日趋激烈,项目环境也日趋复杂,企业对项目管理要求越来越高。为应对复杂环境中出现的各种不确定因素而形成鲁棒性项目调度问题已成为国内外关注的热点。其中,关键链法和资源流网络法的研究最为广泛,此两种方法已在理论和实践中被证明是有效的具有鲁棒性的管理方法。关键链法提出输入缓冲,项目缓冲,资源缓冲及接力赛策略。其中,接力赛策略要求项目各任务节点尽早开工以达到尽快完成项目的目的。与接力赛策略相对应的时刻表策略则要求项目各任务节点不能早于计划时刻开工,此策略通过延长工期来增强项目的鲁棒性。本文将采用关键链法和资源流网络法研究多模式资源受限项目调度问题。在随机DTRTP环境下,按照时刻表策略和接力赛策略模拟仿真执行项目,对项目这两种调度策略进行比较研究,并对模拟仿真得到的数据构建合适的多层次混合模型,采用SAS软件进行影响分析。首先,提出一个识别关键链的启发式算法,通过数值计算说明算法的可行性和有效性。其次,在关键链项目计划中插入缓冲区后,分析可能出现资源冲突和紧前关系冲突,提出基于分支定界法的局部性重排算法和基于分支定界法的全局性重排算法来解决冲突问题。通过模拟仿真,从三个不同层次分析项目任务的不确定性对项目即时完工率和项目惩罚成本的影响,并比较基于分支定界法的局部性重排算法和基于分支定界法的全局性重排算法的优越性。再次,关键链法应用到多模式的随机DTRTP项目调度问题中,从输入缓冲区设置(输入缓冲区设置大小影响)、优先级别(关键链法产生的优先级与其他优先级的影响)和不同层次的可用资源量(可用资源量的影响)三个不同角度,分别按照接力赛策略和时刻表策略进行模拟仿真,对接力赛策略与时刻表策略进行比较分析。最后,资源流网络法应用到多模式的随机DTRTP项目调度问题中,保留较好优先级以及适合随机DTRTP问题的缓冲区大小的关键链法,通过对小规模问题和大规模问题进行模拟仿真来分析优先级及资源流网络各自对接力赛策略和时刻表策略的影响。针对模拟仿真数据,构建符合试验数据的多层次混合模型,采用SAS软件从统计的角度来分析它们的相互影响。更多还原

【Abstract】 With rapid development of economic and market, the environment of project becomes more and more complicate, resulting in high level of project management. In order to deal with various uncertainty in the complicated project environment, robust project scheduling has become one of the most hot and popular topic in the filed of project scheduling probelm. Especially, the Critical Chain Scheduling and Buffer Management (CC/BM) methodology and resource flow network have attack much attention to the world and have been proved to be good methodology to deal with robust project scheduling problem. CC/BM introduced the concepts of feeding buffers, project buffers and resource buffers as well as the roadrunner mentality. This last concept, in which activities are started as soon as possible, was introduced in order to speed up projects by taking advantage of predecessors finishing early on the one side. On the other side, the railway scheduling concept of never starting activities earlier than planned was introduced as a way to increase the stability of the project, typically at the cost of an increase in the expected project makespan. This dissertation applied the CC/BM and resource flow network to the multiple resource constrained project scheduling problem. Railway scheduling and roadrunner scheduling is compared in a stochastic DTRTP environment, a multi-level mixed model is built to fit the data of computational results through SAS, and the impact of resource flow network and priority list on the roadrunner scheduling and railway scheduling is analyzed.Firstly, a heuristic algorithm is proposed to identify the critical chain; some numerical results proved that this algorithm is feasible and effective. Secondly, some resource conflicts or precedence conflicts that might be occure after inserting the feeding buffer are shown and global/local branch and bound rescheduling algorithms are proposed to deal with these conflicts. Some simulations are designed to estimate the three different levels of activity uncertainty on the project completion rate and stability cost. Global branch and bound rescheduling algorithms was proved to be better than the local branch and bound rescheduling algorithms. Thirdly, CC/BM is applied to a stochastic DTRTP problem, three different experiments are designed to compare of roadrunner scheduling and railway scheduling from three different angles:the impact of feeding buffer, the impact of priority lists (critical chain priority list and other priority lists), and the impact of different level of availability. Finally, resource flow network is applied to a stochastic DTRTP environment. Retaining some good priority lists and critical chian priority list with good feeding buffer, computational experiments are simulated on the small-scale and larger-scale instances sets. A multi-level mixed model is designed to fit the data of computational results and some analysis are made for the impact of resource flow network/priority list on the roadrunner scheduling and railway scheduling by using SAS software.更多还原