【作者】 张萌；

【导师】 李国喜；

【作者基本信息】 国防科学技术大学 ， 机械工程， 2013， 博士

【摘要】 大规模定制致力于以大规模生产的效率和成本向市场提供多样化和个性化的产品，已经逐步成为现代制造模式发展的主要趋势。产品配置设计是实现大规模定制的关键使能技术之一，有助于企业提高获取客户订单的效率，缩短产品设计周期，以较低成本迅速满足用户的个性化需求，从而提升企业的响应能力和竞争能力。基于模块化的产品配置设计作为一种主要的配置技术手段，是实现产品快速定制的核心与关键。论文围绕复杂机械产品模块化配置设计目前面临的关键理论和技术问题展开系统而深入的研究，设计了基于模块化的产品配置设计总体方案，重点讨论了模块聚类与优化、产品平台构建、多目标配置优化、配置性能预测和配置功能求解等关键技术，建立了产品模块化配置设计的技术体系结构模型，有助于形成覆盖产品模块化配置设计全过程的技术体系。论文的研究工作及取得的主要成果如下：⑴提出了产品非均匀粒度模块聚类与优化方法，基于相关性分析进行零部件模糊聚类，建立零部件分层递阶结构，将不同粒度层级的模块进行组合并采用改进型非支配排序遗传算法，生成非均匀粒度模块聚类方案的Pareto解集。该方法使得模块划分更具柔性，从而在丰富可行解空间的同时有利于实现划分方案集的全局优化，能够满足客户和企业对各个分目标的不同需求和动态需求，有助于解决模块聚类方案对客户需求和企业侧重点的适应性问题。⑵提出了自底向上的模块化产品平台构建方法，利用产品族与模块特征参数的映射、聚类和匹配分析，从模块化产品族中有效地提取公共平台要素，自底向上地形成产品平台。该方法能够充分基于已有的设计资源形成模块化产品平台，保证了较高的资源重用率，而且一定程度上有助于在面向大规模生产的通用性需求和面向定制生产的个性化需求之间找寻到较好的平衡点，使得建立的产品平台具备良好的代表性和合理性。⑶提出了面向不确定信息的产品多目标配置优化方法，综合考虑产品实际配置信息可能呈现的多种不确定形式，通过对各种不确定配置信息的一致化表达，采用改进型非支配排序遗传算法对多目标配置问题进行求解，得到配置方案Pareto解集。该方法有效弥补了基于确定信息的准确配置方法和基于模糊信息的模糊配置方法的缺陷，进一步拓展了产品配置优化方法的适用范围。⑷提出了基于灰关联分析与支持向量机的配置性能预测方法，通过在有限的历史产品族数据中挖掘模块参数与配置产品性能的非线性关系，建立配置性能预测模型，从而预测新配置产品的性能参数取值。该方法不同于传统的试验方法，而是通过软计算技术评估配置产品的性能参数，降低了产品生产成本，缩短了产品研制周期，并能有效解决预测模型结构复杂和预测误差大等问题，为配置产品性能的准确快速预测提供了新的技术途径。⑸提出了面向产品配置设计的功能求解与建模方法，在考虑合理的映射分层和映射关系的基础上，建立了一种新的多层级混合功能求解模型，并将其应用于产品配置设计的后置处理阶段。建立的功能求解模型能够有效结合设计师的创新思维和计算机的逻辑推理能力，兼顾设计创新与资源继承，能够有效保证配置方案改进设计结果的有效性和合理性，以及客户满意度的最大化。⑹建立了计算机辅助的产品模块化配置设计系统，选择通用三维建模软件作为产品结构设计工具，基于论文的研究成果，设计并开发了计算机辅助的产品配置设计系统。该系统可以有效覆盖产品模块化配置设计过程，能够实现配置设计过程的可视与可控，进而提高产品配置设计的效率和成功率，具备广阔的工程化应用前景。更多还原

【Abstract】 Mass customization, being committed to provide various and individual productsto the market with high efficiency and low cost of mass production, has already been themain trend of modern manufacturing. As one of the key enabling technologies of masscustomization achievement, product configuration design contributes to improve theorder acquisition ratio, shorten the product design cycle, save the cost, and increaseresponse ability and competition ability of firms. Module-based configuration design,regarded as a primary configuration technique, is the core and key of realizing rapidproduct customization. Several academic and technological problems weresystematically and deeply studied in this thesis. The overall scheme of modularconfiguration design was proposed, and some critical technologies such as moduleidentification and optimization, product platform development, multi-objectiveconfiguration optimization, configuration performance prediction, configurationfunction solving, and configuration system development were derived, which form atechnological architecture that could cover the whole configuration design process. Themain achievements are described as follows:(1) A new method for module identifying and optimizing based on unevengranularity was put forward. The parts of a product were fuzzily clustered usingrelativity analysis so as to set up a hierarchical structure. All the universes of thegranular layers in the hierarchical structure were gathered and the uneven granularitybased module clustering scheme was formally presented. Four quantified indices wereproposed to set up four optimization objective functions. Use nondominated sortinggenetic algorithm II to solve the problem in order to obtain the Pareto optimal set. Thisapproach makes module identification more flexible, enlarges the solution space ofmodule clustering, and administers to reach the global optimum solution. Moreover, theproposed method can satisfy diverse and dynamic requirements of customers andcompanies to different objectives, which helps to enhance the adaptability of moduleschemes to customers’ requirements and enterprise emphasis.(2) A bottom-up method for module-based product platforming through mapping,clustering and matching analysis was presented. The framework consisted of three steps:mapping parameters from existing product families to functional modules, clustering themodules within existing module families based on their parameters so as to generatemodule clusters, and selecting the satisfactory module clusters based on commonality,and matching the parameters of the module clusters to the functional modules in orderto capture platform elements. This method ensures the effectiveness of design resourcereusing, and is helpful to obtain a balance between the universal requirements orientedto mass production and individual requirements oriented to personalized customization. The approach could enable the developed product platform be representative andfeasible.(3) A new multi-objective optimization approach to configuration design with theconsideration of several types of uncertain information was brought forward. Theuncertain configuration information was uniformly described with interval numbers. Amulti-objective optimization model was generated by integrating three mathematicalmodels such as the performance, cost and term. The non-dominated sorting geneticalgorithm II was used to solve the model and a Pareto optimal set of productconfiguration schemes was obtained. This method can effectively deal with the problemof configuration optimization under uncertain information, which remedies the defectsof specific configuration and fuzzy configuration. Therefore, the scope of theconfiguration optimization approach could be further extended.(4) A novel prediction approach based on the integration of grey relational analysisand support vector machine through discovering the knowledge from the historicalconfiguration information was proposed. Product performance prediction at the end ofthe configuration process can estimate the performance parameter values through thesoft computing method instead of practical test experiments, which is propitious toreduce the cost, shorten the development cycle, decrease the complexity of predictionmodel and reduce the prediction error. This methodology, ensuring the performanceprediction executed in a short period of time with a high degree of precision, even underthe small sample conditions, leads to a new technique for predicting configurationperformance.(5) An approach to function solving and modeling oriented to configuration designwas raised. Considering rational design domains including function, working principle,behaviour and structure, a hierarchical functional solving model with hybrid mappingswas established, which could be used to address the problem during the post processingstage of configuration design. The model can effectively combine the creative thinkingof designers with the power of reasoning computing of computers, and administers tothe implementation of inheritance and innovation. The application of this functionalsolving model to configuration design could ensure the validity and feasibility of designresults and the maximization of the customers’ satisfaction.(6) A prototype of computer-aided modular configuration design software wasdeveloped and implemented to verify the above theories and technologies. Theprototype was built based on the general CAD platform and consisted of the functionssuch as configuration data management, configuration model establishment,configuration model management and configuration model solving. The preliminaryapplication was carried out to demonstrate the proposed methods. This prototype,covering the whole process of product configuration design, enables to realizesynchronous visualization and complete control of the configuration process, improve the efficiency and success rate of configuration design, and it shows a broad marketableprospect.更多还原