【作者】 叶建芳；

【导师】 潘晓弘；

【作者基本信息】 浙江大学 ， 机械制造及其自动化， 2011， 博士

【摘要】 面向订单装配(assemble to order,ATO)是一种先进的生产组织模式,在这种模式下,企业交货期短,库存水平低,可以提供给客户多样化、定制化的产品,因而在国外企业得到了广泛的应用。但针对我国的具体情况一直缺乏系统的理论指导,在实践中的应用也比较少。本文结合浙江省重大科技攻关资助项目(NO.2003C11010,NO.2005C11034)的研究,在深入分析和讨论电子制造企业的业务流程、生产组织的特点基础上,提出了在计划层面上的产品和零部件备货优化的计划方法和在执行层面的外购件补货控制方法和自制件生产闭环控制方法,为企业应对复杂的生产环境提供了理论和方法指导。第一章回顾了课题研究的背景,阐明了不确定性环境下面向订单装配模式生产计划和控制方法的研究意义。重点论述了面向订单装配、不确定环境下主生产计划以及库存管理、车间作业调度等方面的国内外研究现状。在此基础上,给出了本文的研究目标、研究内容和论文结构。第二章首先电子制造业的特点、ATO系统的运作模式,提出了电子制造企业准ATO生产模式,进一步分析了准ATO模式的实施方法,提出了一个四层的方法体系,分析了其中的关键技术,引出本文所主要解决的三个方面的问题。最后,提出了基于支持向量机的产品市场需求预测方法,并提出用混沌粒子群算法来进行支持向量机参数优选。第三章首先提出了电子制造企业准ATO模式下的主生产计划方法。在此基础上,利用可行性规划的相关理论,建立了一种模糊环境下生产计划优化模型,通过等价转换的方法将上述模型转化为双目标带约束的清晰等价形式。针对该模型提出了一种求解此模型的基于协同进化和不可行度的改进粒子群算法。并进行了实例验证。第四章针对制造车间的诸多因素影响着生产计划在车间的执行的问题,本文提出闭环控制的方法来减少车间不确定性因素对计划执行的影响。该方法分成三个部分：首先根据加工时间分布、机床故障率等随机因素进行随机规划生成初始调度方案,将初始调度方案下到车间,在执行过程中采集过程数据,进行动态事件监测,如有预先定义的动态事件发生则触发动态调度,根据系统当前的状态,找到相对应的调度规则,迅速生成调度方案,减少例外事件的影响。ATO模式下的排产问题是一个多产品、多订单、多资源的组合优化问题。在第五章本文首先应用基于熵权法的模糊综合评价方法进行客户订单优先级的评估,在此基础上,建立了ATO模式总装排产模型,并提出了一种动态改变惯性权重的粒子群算法进行求解。第六章对总结了本文研究所取得的成果,并指出了今后的研究方向。更多还原

【Abstract】 In recent years, assemble-to-order (ATO) is an advanced production organization method.By this method, enterprise can provide customized product with low cost in short delivery time. ATO is widely adopted by abroad enterprise.ATO has great potential in the future. Sponsored by the Key Sci. & Tech. Program of Zhejiang Province, China (No.2003C 11010, NO.2005C11034), based on in-depth analysis of the ATO operation process, this dissertation focused on the production planning and control. That provided a solid theoretical basis and guidance method for the successful implementation of ATO and rapid response to market demand and customer orders.In chapter one, the background and significance of research paper is stated, the research state at home and abroad related to ATO is summarized, and the research objective, and presented the main research content and paper structure is presented.In chapter two, based on the analysis of characteristics of electronic manufacturing enterprise and the analysis of the ATO mode, the quasi-assemble-to-order mode is proposed. The four-layered implementation method system is put forward based on the comparison between ATO and quasi-ATO and three key problems that this thesis focuses on are analyzed in detail. In the end,the demand forecast techniques based on support vector machine (SVM) is proposed and the chaos particle swarm optimization algorithm is put forward to solve the parameter selection problem.In chapter three, a master planning schedule based on storing products and parts for fast delivery was brought forward. A fuzzy production planning model based on credibility programming under fuzzy environment was constructed and transformed into the form of a clear equivalent through the clarity of fuzzy objectives and constraints. A modified PSO algorithm based on coevolutionary and infeasibility was given for solving this model.In chapter four, we focus on the production control of component manufacture.Firstly, since the process time is random, its distribution is calculated by regression method. Secondly, a stochastic programming model is built to find the optimum initial shop floor production planning and an integrated intelligent algorithm that based on Monte-Carlo simulation, particle swarm optimization and SVM is proposed to solve the model. Thirdly, an integrated shop floor dynamic scheduling framework was introduced. This framework includes a scheduling drive mechanism and a dynamic scheduling method which was based on the real-time information of shop floor. Since the scheduling method is based on machine learning, the main problem is scheduling feature selection. Finally we put forward an immune binary particle swarm optimization to select the appropriate features。 Scheduling under ATO environment is a muliti-product,multi-order and multi-resources combinatorial optimization problems.In chapter five, the sequence of scheduling orders was given by calculating order priority value based on evaluation orders priority indicators and fuzzy comprehensive appraisement model which using entropy weight. The assembly scheduling model was established. A new adaptive particle swarm optimization algorithm with dynamically changing inertia weight (DCWPSO) was brought forward to solve the problem.In chapter six, the main conclusions of this dissertation are summarized and the further research issues are put forward.更多还原