| 题名 | 面向混合流水车间的排产优化问题研究 |
| 作者 | 韩忠华 |
| 学位类别 | 博士 |
| 答辩日期 | 2014-05-19 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 导师 | 史海波 |
| 关键词 | 混合流水车间 排产优化 生产事件 概率分布密度 差分进化算法 多目标优化 仿真平台 |
| 其他题名 | Research on Scheduling Optimization Problem for Hybrid Flow Shop |
| 学位专业 | 机械电子工程 |
| 中文摘要 | 混合流水车间具有多道工序、多个加工工位、并且至少有一道工序具有多个并行工位，加工过程存在多条工艺路线，不同工位加工工时又存在差异，这些因素使得混合流水车间排产问题较一般排产问题复杂。装配型生产线以及串并型结构的批制造过程（Batch Process）都是典型的混合流水车间，涉及汽车生产、半导体封装、大型装备制造等诸多国民经济的关键行业。这类生产车间具有工艺复杂、制造过程耦合紧密、各生产环节受到时序约束等特点，对生产管控提出了更高要求。车间级排产作为制造执行系统的重要部分，目前在解决混合流水车间排产问题时，难以同时满足多个优化目标需求，评价指标体系不够完善，排产算法的优化效果不够理想等问题，导致排产结果不能有效地指导这类车间的生产运作过程，所以研究混合流水车间排产优化问题具有重要的理论意义和实际应用价值。 本文针对混合流水车间的排产优化问题，从建立混合流水车间数学规划模型入手，分析混合流水车间动态运行规律，然后从混合流水车间排产优化问题的优化方法和优化目标两个方面进行深入研究。在优化方法方面，通过研究初始种群的构建方法和改进群体智能优化算法，来找到更好的求解混合流水车间排产优化问题的方法。在优化目标方面，分析多个具有混合流水车间特点的优化目标，建立评价指标体系，并分析比较多种群体智能算法针对不同优化目标的排产优化效果。在此基础上研究了混合流水车间多目标优化问题，从多方面改进多目标优化算法，进一步增强混合流水车间排产优化方法的实用性和有效性。 本文主要工作分为以下几个方面： (1)建立混合流水车间排产优化的数学规划模型，其涵盖了混合流水车间中的工位、工序、流程等对象，也包括了这些对象之间的约束关系。建立反映生产过程的事件模型，而排产过程又是对生产过程的一种模拟，在事件模型中研究随时间变化的生产状态迁移的影响和工序对排产过程的拉动作用，进而更深入解析混合流水车间的排产过程。 (2)建立初始种群是群体智能进化算法解决混合流水车间排产优化问题的关键。本文中提出的多种初始种群构造方法都是基于工序变工位的初始种群建立方法，将基于分布概率的初始种群构造方法与基于优化目标的初始种群构造方法相结合，提出了基于联合选择概率的初始种群构造方法。该方法能够反映优化目标的特点，控制种群中个体在基因片段上的分布性。并通过基于信息熵的方法控制种群中个体在解空间上的分布性。综上所述，为了达到提高初始种群中生成个体的质量和提高生成个体在解空间的分布性的目的，提出基于信息熵的联合选择概率初始种群构造方法。 (3)提出一种动态双种群自适应差分进化算法（DDPSADE）用于求解混合流水车间排产优化问题。该算法在标准差分进化算法基础上，利用全新的双种群独立进化与交流进化机制，既可以巩固优秀个体的进化趋势，也能够增强进化的活力；利用重新设计的自适应改变进化算法参数的方法，在进化过程中加入扰动，增强跃出局部极值的能力；利用加入动态更新高相似性个体的方法，来减少冗余运算和扩大解空间的搜索区域，从而达到全面提高优化效果的目的。并通过与遗传算法（GA）、粒子群算法（PSO）、差分进化算法（DE）、粒子群和差分进化混合算法（PSDEHA）等多种算法的对比分析来验证该算法的优化性能。 (4)针对单目标排产优化问题，首先分析和描述这些混合流水车间排产优化目标，建立评价指标体系，评价分析上述各种群体进化算法在求解混合流水车间排产优化问题的效果。在系统的研究单目标问题基础上，进一步研究混合流水车间的多目标排产优化问题，通过分析NSGA-II算法，将求解多目标优化问题归结为6个主要环节，分别进行改进和探索，特别是在分布性方面，首次提出采用了AP聚类方法进行分布性研究，在进化计算部分，也提出多目标动态自适应差分进化算法，通过采用基于AP聚类的多目标动态自适应差分进化算法有效提高混合流水车间多目标排产优化效果。 (5)开发混合流水车间排产优化仿真平台，设计的系统架构能够满足面向网络化应用的需求，重点研究生产线动态建模、排产优化驱动引擎和企业消息总线三个部分。 |
| 索取号 | O224/H15/2014 |
| 英文摘要 | In Hybrid Flow Shop (HFS)，there exists multi-stage, multi-machine, and at least one stage existing multi parallel machine, production process existing multi process routes, and the operation time varying from machine to machine, all these factors make the scheduling problem of HFS is more complicated than the general scheduling problem. Both the assembling production line and the Batch Process (BP) which is series parallel structure are typical HFS, which includes car manufacturing, semiconductor packaging, large-scale equipment and other key industries of national economy. Due to the characteristics of complicated technology, tightly coupled manufacturing process, and constrained by time sequence in each production link in this kind of workshop, it will put forward higher request of production management and control. As the key part of Manufacturing Execution System (MES), workshop scheduling has the following problems in solving the HFSP currently, such as the multi optimal objective cannot be satisfied, the evaluating indicator system is not perfect, and the optimization effect of the scheduling algorithms is non-ideal, which mean that the scheduling result cannot direct the production process of this kind of workshop effectively. So it is of great both theoretical and practical significance that make research on Scheduling Optimization Problem (SOP) of HFS. Aiming at the HFS scheduling optimization problem, this paper starts with modeling the mathematical programming model, then analyzes the dynamic operation rules of HFS, makes deep research on two aspect, optimization methods and optimization goals. In optimization method, it is seek out better methods to solve HFS’s scheduling optimization problems through studying the construction methods of the initial population and improving swarm intelligent optimization algorithms. In optimization objectives, the multi optimization objectives with HFS characteristics are analyzed, the evaluating indicator system is built, and the scheduling optimization effects on different optimization objectives are compared among multi swarm intelligent algorithms. On the basis, the multi-objective optimization problem of HFS is studied, and the multi-objective optimization method is improved from multi aspects, which enforces the practicability and effectiveness of scheduling optimization method of HFS. The main works of this paper includes the following parts: Firstly, the mathematical programming model of HFS scheduling optimization problem is modeled, which includes the objects of machines, stages and process in HSF, also includes the constraint relationship between these objects. The event model reflecting the production process is modeled, while the scheduling process is a simulation of production process. In event model, the effect of the production state transition varying with the time, and the pull effect of stage to scheduling process are studied, then the scheduling process of HFS are analyzed more deeply. Secondly, the key to solve SOP of HFS using swarm intelligence evolutionary algorithm is establishing initial population. In this paper, the proposed initial population construction methods are all based on the idea of stages changed into machines (SCM). Combining the distribution probability based and optimal objective based initial population construction method, a joint selection probability based initial population construction method is proposed. This method can reflect the characteristic of the optimal objective, control the distribution of individual in gene segment, and control the distribution of individual in solution space through information entropy method. Above all, in order to improve the quality of individuals in initial population and improve the distribution of individual in solution space, a information entropy based joint selection probability method to construct initial population is proposed. Thirdly, a Dynamic Double Population Self-adaptive Differential Evolution algorithm (DDPSADE) is proposed to solve the SOP of HSF. This method based on the standard Differential Evolution (DE) algorithm, can not only consolidate the evolutionary trend of the excellent individuals but also can enhance the evolutionary energy by using the brand new double population independent evolution and exchange evolution mechanism. Using the redesigned method of changing the evolution parameters self-adaptively, putting disturbance into evolution process, so that the ability to jump out local extreme value can be enhanced. Using the method of dynamic updating high similarity individuals to decrease the redundant operations and expand the search region of solution space, so that the optimization effect can be improved comprehensively. Finally, the comparison and analysis among Genetic Algorithm (GA), Particle Swarm Optimization Algorithm (PSO), Differential Evolution Algorithm (DE), Particle Swarm and Differential Evolutionary Hybrid Algorithm (PSDEHA) are made to testify the optimal performance of the proposed method. Fourthly, for the SOP with single objective, the objective of SOP in HFS is analyzed and described firstly, the evaluating indicator system is established, and the effects of SOP in HFS using all kinds of swarm evolutionary algorithms are evaluated and analyzed. On the basis of studying the single objective systematically, the multi-objective of SOP in HFS is studied fatherly. By analyzing NSGA-II algorithm, the solution seeking of the multi-objective optimization problem can come down to 6 main parts improved and explored respectively. Especially in the part of distribution, AP clustering method is firstly used to study the distribution. In the part of evolutionary computing, multi-objective dynamic self-adaptive differential evolutionary algorithm is also proposed, and the effect of the multi-objective scheduling optimization of HSF can be improved through adopting the AP based multi-objective dynamic self-adaptive differential evolutionary algorithm. Fifthly, the simulation platform of SOP in HFS is developed. The designed system construct can meet the demand of facing to internet application. The production line dynamic modeling, scheduling optimization driving engine, and enterprise message bus are studied emphatically. |
| 语种 | 中文 |
| 产权排序 | 1 |
| 页码 | 125页 |
| 分类号 | O224 |
| 内容类型 | 学位论文 |
| 源URL | [http://ir.sia.ac.cn/handle/173321/14789]  |
| 专题 | 沈阳自动化研究所_数字工厂研究室 |
| 推荐引用方式 GB/T 7714 | 韩忠华. 面向混合流水车间的排产优化问题研究[D]. 中国科学院沈阳自动化研究所. 2014. |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
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。
修改评论