【作者】 黄亮；

【导师】 王德伦；

【作者基本信息】 大连理工大学 ， 工业工程， 2010， 博士

【摘要】 本文基于国家自然科学基金项目(No.61074135, No.50275016, No.50475154和No.50775016)中的特征状态空间理论及方法,以机械制造企业的多阶段混流生产过程为研究对象,提出并建立了生产过程的特征状态建模、分析与控制方法。以冰山集团大连冷冻机厂、大连重工·起重集团等机械制造企业的生产过程与数据为依据,通过实际算例验证了本文提出的理论与方法。首先,本文对机械制造企业中的各段局部生产过程进行建模,定义了基本生产活动。以基本生产活动为基本分析单元,提取物料以及其它生产资源流经基本分析单元前后的时间、质量和成本等特征信息,形成特征状态矢量。根据特征状态矢量的变化规律构建基本生产活动的特征状态方程,以特征变换矩阵的形式对基本生产活动的执行情况进行直观、定量地表达。其次,本文建立了机械制造企业整体生产过程模型。在基本生产活动特征状态方程表达基础上,针对机械制造企业中普遍存在多阶段混流生产过程,定义物料供应和设备共享两种基本生产活动之间的衔接关系。进一步以特征状态矢量的乘法与加法运算分别表达基本生产活动的串联与并联关系,根据多阶段混流生产过程中的物料流和能量流,形成了多阶段混流生产过程的特征状态方程模型。讨论了特征状态方程在计算机系统中的实现方法。再次,本文对生产过程模型的几何意义以及与之相关的计算性质进行了讨论。定义机械制造系统的特征状态空间,讨论了空间性质。利用特征状态方程将生产过程映射到特征状态空间中研究,将机械制造系统的瓶颈分析和生产控制问题分别转化为特征状态空间中的梯度分析和路径规划问题,为这些问题的解决提供了新的方法与相应的理论依据。然后,基于所建立的生产过程模型及其性质,介绍生产过程简化、分析与控制的具体方法,并通过算例进行验证。利用特征状态方程的符号计算实现了生产过程关键路径提取,简化了生产过程的特征状态方程描述；对简化后的方程求导,实现了制造系统瓶颈的定量分析；将瓶颈分析结果作为中长期生产控制问题的目标函数梯度,将梯度搜索策略集成进入邻域搜索算法,加速算法收敛,减少中长期生产控制问题求解时间。利用多个来源于合作企业的实际算例和仿照实际问题构造的虚拟算例,分别从测试生产过程模型精度、瓶颈分析方法性能和中长期生产控制方案优选算法性能多个方面验证了本文所提出理论与方法的可行性和有效性。与同类研究所使用的其它代表性方法对比,新方法明显提高了几种典型中长期生产控制问题的求解速度和质量,为机械制造系统的分析与控制探寻了一种新方法,开辟了一条新途径。最后,利用本文提出的理论与方法,在为大连冷冻机厂、大连重工·起重集团等企业开发生产管理信息系统的基础上,设计并开发了网上看板控制系统,实现了生产过程的状态监测、瓶颈分析和几种典型生产控制问题的计算机辅助求解。更多还原

【Abstract】 A new methodology for modeling, analyzing and controlling the mutil-stage and mutil-product production processes in mechanical manufacturing enterprises is developed, which is based on the characteristic state space method supported by the National Natural Science Foundation of China under Grant No.61074135, No.50775016, No.50475154 and No. 50775016. According to the production processes and the production data of Bingshan Group Dalian Refrigeration Co. and DHI-DCW Group Co., a number of case studies were provided to reveal the feasibility and validity of the new methodology.First, local production process models are set up. Basic production activities are regarded as basic analysis units after they are defined. Characteristic state vectors are established using the variables for representing the time, the quality and the cost of the materials and the other production resources before and after those flow though the basic analysis units. Characteristic state equations are set up according to the changes of the characteristic state vectors. Then, the performance of the basic production activities are described intuitively and quantificationally in terms of characteristic transformation matrices.Second, global production process models are set up. Two relationships of the basic production activities, which are called supplying materials and sharing capacity respectively, are defined for the mutil-stage and mutil-product production processes in mechanical manufacturing systems. The sequential relationships and the parallel relationships of the basic production activities are represented by multiplication and addition of the characteristic state vectors respectively. Then, the model of the mutil-stage and mutil-product production processes in terms of characteristic state equations is established according to the materials flows and the energy flows in the processes. The methodology for implementing the characteristic state equations on computer systems is also discussed.Third, the geometric meaning and the associated properties for computing of the proposed production process model are discussed. The characteristic state space of a mechanical manufacturing system is defined and its properties are discussed. Based on these properties, the production processes are studied in the characteristic state space and the bottleneck analysis and the production control of the mechanical manufacturing systems are considered as gradient analysis and path-planning in the space, respectively. This study provides a new way for the bottleneck analysis and the production control and presented the relevant theories on which they are based. Forth, based on proposed production process model and its properties, the approaches of simplifying, analyzing and controlling production processes are introduced in details. The production processes are described by simplified characteristic state equations by finding the critical paths using the symbolic computation of the equations. The bottlenecks of the mechanical manufacturing systems are measured quantificationally using the derivatives of the simplified equations. The measures of the bottlenecks are also the gradients of the object functions of the medium to long term production control problems. Consequently, the gradients can be used to guide the neighborhood-generation in a neighborhood search algorithm in order to accelerate convergence and short the run time of the neighborhood search procedure. A number of actual cases and hypothetical cases are computed for testing the production process model, the proposed bottleneck analysis method, and the modified neighborhood search algorithm. Compared to the classical methods, the computational experiments show relatively positive results. These studies pave a new way for analyzing and controlling the Production Processes in Mechanical Manufacturing Enterprises.Finally, a web-based kanban system is designed and developed using the proposed theories and methods, based on the production management information systems of the Bingshan Group Dalian Refrigeration Co. and DHI-DCW Group Co. The production processes are monitored, the bottlenecks of the manufacturing systems are detected and measured, and several typical production control problems are solved by means of the web-based kanban system.更多还原