【作者】 贡智兵；

【导师】 李东波；

【作者基本信息】 南京理工大学 ， 机械制造及其自动化， 2007， 博士

【摘要】 随着社会和科学技术的发展市场竞争的日益加剧，客户对产品交货期、价格、质量和个性化的要求越来越高。大规模定制生产模式(Mass Customization，MC)把产品定制和产品大批量生产结合起来，在降低成本、保证质量的同时，以最短的时间满足客户个性化需求。面向大规模定制设计(Design for Mass Customization，DFMC)是MC生产模式的关键技术之一。基于产品平台(Product Platform，PP)的产品配置设计是DFMC的重要方法，是MC产品开发的研究热点。本文围绕基于产品平台的快速配置设计流程展开研究，研究的技术包括模块划分、产品平台构建与评价、产品配置方法、配置产品变型设计规划、结构的快速优化。主要研究内容如下：首先，研究了面向配置设计的模块形成过程及划分方法。通过分析客户需求和零部件相关性，建立零部件模糊相似性矩阵，并构建零部件模糊聚类树来描述模块的形成过程。提出采用复合λ对模糊聚类树截距，以组合复杂度、设计复杂度和成本为优化目标进行模块划分。利用遗传算法进行模块划分的优化，得到各λ值，进而得到模块划分的优化结果。此方法不仅考虑了零部件之间关联的紧密程度，而且考虑了模块划分的目标。以某天文圆顶为例说明方法的可行性和适用性。以模块划分为基础，研究了产品平台构成、构建原理和评价方法。提出模块通用性的量化准则和算法，以模块通用性为优化目标，以复合λ值模糊聚类树为基础，进行模块的聚类。提出并构建产品平台的评价体系，分析模块通用性和其它评价指标之间的关系，采用分级优化方法对产品平台进行优化。以产品平台为基础，构建了产品优化配置模型。从客户需求出发，建立以产品性能和价格为目标的多目标优化模型，并运用PGA进行方案的配置和选择。变型设计是产品配置的重要环节，由于变型设计中结构参数的多样化和耦合性，设计过程需要多次迭代。为了减少设计的迭代次数和设计的风险，增加一次设计成功的概率，提出基于设计结构矩阵(Design Structure Matrix，DSM)的实例变型路径规划。本文研究了产品变型设计的活动规划必要性和传统DSM的不足，提出把传统DSM扩展为静态DSM和动态DSM进行变型设计的路径规划，提高了变型设计的一次成功率。另外，在分析了以有限元为基础的变量化设计和优化的不足基础上，提出用BP神经网络和GA结合的快速分析和优化策略。通过实例对比两种方法优化效率和优化结果，结论表明本文的分析和优化策略不仅效率高而且有利于程序的控制。最后，结合南京中科天文仪器公司的工程项目，基于SolidWorks开发了天文望远镜的快速设计系统，该系统在实践中得到初步应用，效果良好。更多还原

【Abstract】 With the development of the society and the scientific technology, the enterprises arefacing increased competition from the market. The customers’ requirements, such asproducts lead time, products cost, products quality and individuation, are higher and higher.Mass customization (MC), combining product customization and batch production, cansatisfy customers’ individuation with low cost and high quality of products. Design formass customization (DFMC) is one of the key technologies of MC.Product configuration design based on the product platform (PP) is the primarymethod and the research focus of MC. The process of the rapid product configurationdesign based on the PP is studied in detail in this paper. The technologies include: modulepartition, PP establishment and evaluation, product configuration methods, variant designplanning and rapid structure optimization. The main research contents in this paper are asfollows:Firstly, module generation and partition methods for product configuration are studied.After analyzing customer demands and the correlativity of parts, the fuzzy similaritymatrix and fuzzy cluster tree are constructed to illustrate the process of module generationin product configuration. Then, the module partition method with three optimizationobjects(module assembly complexity, design complexity and cost) based on the fuzzycluster tree intercepted by combinedλis put forward. Using GA to optimizeλvalue ofeach module to achieve the optimization module partition. The method considers not onlythe correlativity of parts, but also the target of the module partition. The example ofastronomical dome module partition shows the method in this paper is feasible andapplicable.Secondly, the methods of PP establishment and evaluation are researched based onmodel generation and partition. The quantization criterion and computation algorithm ofmodule universality are put forward. Based on combinedλfuzzy cluster tree, the modelsare clustered aiming at module universality optimization. The PP evaluation system isestablished. Based on the proposed evaluation system, the relation between modelscommonality and each evaluation index are analyzed, and the PP is optimized by two-stepoptimization method. The model for product configuration design is built on the basis ofthe optimized PP. According to the customer requirements, the Multi-Objective Optimization (MO) model is built regarding capability and cost as optimization targets.With the optimization model, the configuration and choice of product scheme is completedby using PGA.Thirdly, variant design, which is the key technology of product configuration becauseof the diversity and coupling of the design parameters resulting in iterative variant designprocess, are studied. The variant design path planning by using Design Structure Matrix(DSM) is put forward to reduce the iterative times, the design risk, and increase the singlesuccess ratio. In this paper, after analyzing the necessity of the variant design path planningand the shortcoming of the traditional DSM, the extended DSM with dynamic DSM andstatic DSM is put forward to plan the variant design path.After the analyses of the shortcoming of the variantional design and optimizationbased on the FEA, the rapid analysis and optimization strategy using BP artificial neuralnetwork and Genetic Algorithms (GA) is put forward. The two optimization methods iscontrasted by an example, and the results show the methods in this paper is much moreeffective and controllable than that of the FEA.Finally, the Rapid Design System of Astronomical telescope of the project in NanjingAstronomical Instruments Co., Ltd was developed based on SolidWorks. The system wasapplied in some telescope design. The result shows that the system is valuable.更多还原