【作者】 张炜；

【导师】 侯亮；

【作者基本信息】 厦门大学 ， 车辆工程， 2021， 博士

【摘要】 大规模定制战略背景下,单个企业所提供的有限产品难以实时响应日益增长的个性化需求。而云制造可充分利用分布式企业间制造资源的横向集成、单个企业内混装线系统的纵向集成,实现小批量、多样化用户需求,得到广泛青睐。然而,云制造涉及动态变化的个性化需求、分布式企业间制造资源以及企业内混装线系统三个层级的利益主体,且各层级内也存在利益冲突,因而加剧了制造系统内利益主体的关联和博弈。同时,云制造产生了多源实时数据且其价值常被忽略,如何有效挖掘这些数据并反馈给云制造系统各主体,实现面向个性化需求的多层级制造资源协同及产线优化,亦是云制造背景下急需解决的关键问题。针对上述问题,本文在分析多层级各利益主体耦合关系的基础上,提出了基于数据驱动的云制造系统反向设计理论与方法;对云制造背景下的企业间制造资源横向集成、企业内混装线系统纵向集成进行建模;通过加强分布式制造资源之间以及混装线系统内部的协同,适应动态变化的个性化需求。论文研究内容主要包括以下几个方面:定义了一个理想化云制造系统架构,并分析基于云平台制造资源配置服务一般流程,开展数据资源特征及需求分析。提出数据反向驱动的云制造设计决策框架,通过系统实际状态数据的挖掘和学习来配置个性化需求。从博弈论的角度分析工程设计领域决策问题内多主体、多目标、多约束的耦合以及层次性优化关系,并建立具有正反向交互特征的主从协同优化模型。引入了主从协同优化决策模式以改进传统遗传算法,用以平衡上层模型和下层模型之间的冲突。针对个性化需求和分布式资源层次协同关系不明确问题,基于制造系统反向设计方法,提出了基于云平台数据驱动产品需求和制造资源主从协同决策框架。正向客户需求的产品配置充当上层主者,下层供应商联合配置充当从者反向影响产品配置的决策。嵌套遗传算法用以进行模型权衡解优化,并通过实例验证方法和算法的有效性。考虑订单分解和投产排序协同时多条混装线存在能力、作业时间不等效的特征,提出面向柔性定制的并行不等效混装线生产计划模型。分析两者之间相互影响和制约的关联机理,建立主从协同优化模型。针对模型特征提出一种结合Pareto前沿解的嵌套遗传算法。引入某客车混装线进行实例分析,验证模型表征企业实际情况的能力。分析投产排序和产线平衡协同时采用固定节拍所忽略的数据实时性问题,以混装线实时状态工作站工时极值为动态节拍。以静态平衡和动态平衡为目标建立联合决策模型。提出改进人工蜂群算法,用以求解平衡和排序的联合优化。在案例集中通过评价指标验证算法有效性,以及动态节拍合理性。基于当前装配序列规划和装配线平衡集成方法存在局限于一种产品的问题;在分析两者非合作博弈决策特征的基础上,提出了混装线装配序列规划和混装线平衡的联合博弈模型。研究博弈主体之间的层次关系,提出了系统的分析和决策机制。利用嵌套遗传算法进行模型优化求解。引入一组经典案例验证所提主从协同优化模型和算法的有效性。基于数据反向驱动技术捕捉了制造系统演进基本规律,实现对制造系统动态调整提供精准化管理,对个性化定制企业具有重要现实意义。混装线系统联合决策过程是一个包含多对矛盾问题、动态平衡的过程。通过研究和掌握其基本规律,建立制造系统联合决策模型,发展了混装线制造系统演进理论体系。更多还原

【Abstract】 With the background of mass customization strategy,the limited products provided by a single enterprise are difficult to respond to the growing personalized demand.Cloud manufacturing can make full use of the horizontal integration of manufacturing resources between distributed enterprises and the vertical integration of the mixed-model assembly line(MMAL)system within a single enterprise to achieve small batch and diversified customers’ needs.However,cloud manufacturing involves three levels of stakeholders:dynamic personalized demand,distributed manufacturing resources among enterprises and MMAL within enterprises.There are conflicts of interest in each level,which aggravates the association and game of stakeholders in the manufacturing system.At the same time,cloud manufacturing produces multi-source real-time data,and its value is often ignored.How to effectively mine these data and feed them back to the main bodies of the cloud manufacturing system to achieve multilevel manufacturing resource collaboration and production line optimization for individual needs is also a key issue that needs to be solved urgently in the context of cloud manufacturing.In response to the above problems,this paper proposes a data-driven cloud manufacturing system inverse design theory and method based on the analysis of the multi-level coupling relationship between stakeholders;modeling the horizontal integration of manufacturing resources between enterprises and the vertical integration of MMAL within enterprises under the background of cloud manufacturing;strengthening the collaboration between distributed manufacturing resources and the internal of MMAL to adapt to the dynamic changes of individual needs.The research contents of this paper mainly include the following aspects:This paper defines an idealized cloud manufacturing system architecture,analyzes the general process of manufacturing resource allocation service based on cloud platform,and carries out data resource characteristics and demand analysis.A datadriven cloud manufacturing inverse design decision-making framework is proposed,and personalized requirements are configured through the mining and learning of actual system state data.From the perspective of game theory,analyzed the coupling of multiagent,multi-objective,multi constraint and hierarchical optimization relationship in decision-making problems in engineering design field,and establishes a leaderfollower joint optimization(LFJO)model with positive and negative interaction characteristics.The LFJO decision model is introduced to improve the traditional genetic algorithm to balance the conflict between the upper model and the lower model.Aiming at the problem of the unclear relationship between personalized demand and distributed resource level coordination,based on the reverse design method of manufacturing system,a LFJO decision-making framework based on cloud platform data-driven product requirements and manufacturing resources is proposed.The product configuration with positive customer demand acts as the leader,and the lower supplier joint configuration acts as the follower,which affects the decision of product configuration.Nested genetic algorithm is used to optimize the trade-off solution of the model,and the effectiveness of the method and algorithm is verified by an example.In order to solve the problem that capacity and operation time of multiple MMAL are not equivalent when order decomposition and production scheduling are coordinated,a parallel production planning model for flexible customization is proposed.The mechanism of interaction and restriction between them is analyzed,and the LFJO model is established.According to the characteristics of the model,a nested genetic algorithm combined with Pareto front solution is proposed.An example of a bus MMAL is introduced to verify the ability of the model to represent the actual situation of the enterprise.Analyze the real-time data problem that is ignored by the fixed beat when the sequencing and balancing are coordinated,and set the extreme value of the working hours of the real-time state workstation of the MMAL as the dynamic beat.Set up a joint decision-making model with static balance and dynamic balance as the goal.An improved artificial bee colony algorithm(IABC)is proposed to solve the joint optimization of balancing and sequencing.In the case set,the effectiveness of the algorithm and the rationality of the dynamic beat are verified by the evaluation index.As for the problem that the current integration method of assembly sequence planning and assembly line balancing is limited to one product,a joint game model of assembly sequence planning and assembly line balancing for MMAL is proposed based on the analysis of their non cooperative game decision characteristics.This paper studies the hierarchical relationship between game players,and puts forward a systematic analysis and decision-making mechanism.The nested genetic algorithm is used to optimize the model.A group of classic cases are introduced to verify the effectiveness of the proposed LFJO model and algorithm.Data-based inverse driven technology captures the basic evolution rule of the manufacturing system,and realizes precise management of the dynamic adjustment of the manufacturing system,which has important practical significance for personalized customization enterprises.The joint decision-making process of the MMAL manufacturing system is a process that contains multiple pairs of contradictory problems and dynamic balancing.By studying and mastering its basic rules,a joint decision-making model of manufacturing system is established,and the evolutionary theory system of MMAL manufacturing system is developed.更多还原