CASIA OpenIR  > 毕业生  > 博士学位论文
可重入生产系统的调度优化与性能分析
赵丽娜
学位类型工学博士
导师郑应平
1999-12-01
学位授予单位中国科学院自动化研究所
学位授予地点中国科学院自动化研究所
学位专业控制理论与控制工程
摘要本文是在国家自然科学基金重点项目“复杂系统建模理论与方法学”的资 助下完成的。 可重入生产系统是以VLSI为代表的一类复杂生产系统。随着微电子行业的 飞速发展(每1~2年翻一番),已受到工业界和学术界的广泛关注。本文对控制与 调度策略,建模方法以及性能分析,检测站设置等方面进行了系统的研究。提出 了一些新策略,新方法,仿真和计算表明了这些策略与方法的有效性和优越性。 主要创新成果如下: 1.针对闭环可重入生产系统的调度,设计了并行顺流定级法(Parallel priority ordered downstream,PPOD)。适用于闭环确定路径系统的调度。与调度效果较好 的负载平衡策略比较具有更好的适应性。 2.针对开环可重入生产系统的调度,提出了可附加于调度策略的储备调度方 法。适当选择最小储备量,可以缩短工件循环时间。 3.将拟生灭过程(QBD)型模型引入可重入生产系统的研究。假设系统中除第 一个外,其余缓冲区容量都有限。输入流是泊松流,加工时间服从负指数分布。 建立了两加工站四缓冲区可重入生产系统的QBD型模型。 4.给出了两加工站四缓冲区系统在RS和LBFS两种调度策略下的随机稳定 的充要条件。对状态转移阵可约的系统,将系统状态集化成不可约吸收的状态集 和可约状态集之和,并分析了不可约的吸收集的稳态性能。 5.用QBD模型,得到了开环系统稳态分布的数值解。并比较了随机调度RS 和LBFS调度策略对不同系统的作用效果。结果表明,调度策略对系统的调度效 果不仅与调度策略有关,还与系统自身的结构有关。与以往的仿真或计算性能指 标边界的方法相比,QBD模型法显然更具优越性。 6.对近似满足乘积形式解的可重入生产系统,以均值分析作为性能分析的工 具,用模拟退火方法解决了检测站的设置的寻优问题。可以在有限时间内搜索到 近优解甚至最优解,节省计算量,而且精度较好。
其他摘要This dissertation is funded by the chief project of National Natural Science Foundation Complex System Modeling: Theory and Methodology. Re-entrant line is a class of complex manufacturing system with VLSI being its typical. With the development of micro-electric techniques, it calls wide attention of both industry and academia. Control and scheduling, modeling and performance analysis together with inspection allocation of re-entrant lines are systematicly studied. New scheduling policies and modeling methods are proposed. Simulation and computation results show their effectiveness and superiority. Contributions of the dissertation are as follows: 1.PPOD(Parallel priority ordered downstream) policy is designed for closed re- entrant lines. It is suitable for closed-loop system with determined route and it has better adaptability than H-W policy, which proved to be superior than the others. 2.Saving scheduling method is proposed for open-loop systems that can be attached to scheduling policies. Suitably choice of minimum stocks can reduce the cycle time of parts. 3.QBD type model is introduced to model re-entrant lines. Assume that all but the first buffer capacities are finite. The input is poisson flow and serving times are exponential. QBD type models are built for systems with two stations and four buffers. 4.Sufficient and necessary conditions of stability are provided for the systems with two stations and four buffers under RS and-LBFS policy. For system whose state transition matrix is reducible, the set of states is divided into absorbing irreducible subset and reducible subset. Then only the former needs to be considered in static performance analysis. 5.Numerical results of static performances of open systems are worked out. RS and LBFS are compared and the results show that scheduling results are affected not only by policies but also by the structure of the system. Comparing with the simulation methods and bound computing methods, QBD model is obviously superior. 6.To re-entrant lines with product form solutions, which can be analyzed by MVA, simulated annealing(SA) is adopted to search optimal solution of inspection allocation. SA can find near optimal or optimal solutions in finite time with less computation efforts but good precision.
馆藏号XWLW536
其他标识符536
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/5706
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
赵丽娜. 可重入生产系统的调度优化与性能分析[D]. 中国科学院自动化研究所. 中国科学院自动化研究所,1999.
条目包含的文件
条目无相关文件。
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[赵丽娜]的文章
百度学术
百度学术中相似的文章
[赵丽娜]的文章
必应学术
必应学术中相似的文章
[赵丽娜]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。