【作者】 苑明海；

【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， 2008， 博士

【摘要】 随着社会和科学技术的发展,现代制造业正面临不可预测、快速多变和不断增强的市场竞争等一系列问题。在这种背景下,可重构制造系统成为了全球经济一体化时代最有竞争力和发展潜力的制造模式。可重构装配线作为可以适应多品种、变批量、多功能、快速交货和短的产品市场寿命生产模式的新型可重构制造系统,也成为近年来装配领域的研究重点。可重构装配线的设计与管理作为可重构制造系统级研究的关键内容,是引导可重构制造系统深入研究和广泛应用的重要途径,通过可重构装配线管理系统的开发,可以为企业实现装配线的重构、平衡、调度及评价提供一个很好的控制与管理平台,增强企业的灵活性和适应性,提高企业的竞争能力。因此,针对可重构装配线管理系统的关键技术展开研究具有重要的理论意义及工程应用价值。本论文研究了基于多Agent的可重构装配线体系结构及其装配线管理技术,主要包括如下几方面:(1)研究了可重构装配线管理系统体系结构。在对可重构装配线系统需求进行分析的基础上,提出了基于多Agent的可重构装配线管理系统体系结构。分析了该体系结构的主要特征,设计了系统的功能模块。最后分析了系统实现的关键技术,为论文的研究确定了方向。(2)研究了可重构装配线建模技术。针对可重构装配线所具有的复杂性、重构性、智能性及自治性的特点,提出了面向Agent的赋时着色Petri网的可重构装配线建模方法。在对装配资源进行分类的基础上,将装配资源Agent内部行为进行封装,建立了装配资源Agent的ATCPN模型,并通过过渡变迁合成可重构装配单元ATCPN模型。根据每个可重构装配单元之间的逻辑关系,确定整个系统的消息传递关系,在此基础上实现了装配线快速重构。分析了可重构装配线的ATCPN模型的性能指标,并提出了AUML模型向ATCPN模型的映射规则及基于ATCPN的交互协议建模方法,通过引入相应的多Agent交互协议模型可以解决模型中存在的死锁和冲突。(3)研究了可重构装配线平衡。分析了可重构装配线的特点,综合考虑工作站的数量、工作站的负荷及装配线效率三个因素,提出了可重构装配线的平衡模型,同时提出了一种优化装配线平衡的混合遗传算法。该算法将模拟退火算法和遗传算法相结合,并采用了交叉概率和变异概率的自适应重构策略,有效避免了算法的早熟,增强了算法全局寻优能力。(4)研究了可重构装配线调度。综合考虑影响可重构装配线优化调度的三个主要因素,即最小化空闲和未完工作业量、均衡零部件的使用速率及装配线调整成本,提出了可重构装配线多目标优化调度的数学模型。同时,提出了一种基于Pareto多目标遗传算法的可重构装配线优化调度方法,该算法在遗传操作中综合运用了群体排序技术、小生境技术、Pareto解集过滤器及精英保留策略,并采用了交叉概率和变异概率的自适应重构策略,保证了群体的多样性,避免了算法早熟,增强了算法全局寻优能力。(5)研究了可重构装配线多目标调度决策。基于改进的层次分析法和信息熵法,综合了主观赋权法和客观赋权法的优势,提出了一种新的确定指标权重值的非线性目标规划模型。同时针对可重构装配线多目标调度方案优选中存在的诸多灰色信息,提出了一种基于灰关联分析的可重构装配线多目标调度决策模型,并利用决策向量与正理想参考向量关联度的贴近度作为决策方案的评判依据,使评价结果更为可靠、可信。(6)研究了可重构装配线评价体系。在分析了可重构装配线评价指标选取原则的基础上,根据可重构装配线的特点和要求,从工程实际出发,对具有多个不确定因素的多目标、多层次综合评价模糊决策进行研究,构建了一个较为系统、科学、全面的可重构装配线的三层指标评价体系,并分析了各个指标的内涵。利用组合赋权法来确定各评价指标的权重,通过最小二乘原理确定评价对象相对于最优参考值的优良隶属度,以此作为评价准则,并结合层次分析法和模糊数学的方法实现对各指标的评价。(7)建立了基于CORBA的可重构装配线管理系统集成框架,研究了装配资源的信息模型,并设计了系统的Agent结构及各Agent之间的通信机制。同时,开发了可视化的ATCPN仿真建模工具,用于可重构装配线的ATCPN建模和性能分析。在上述研究的基础上,开发可重构装配线管理原型系统,该系统已在某特种车辆厂中得到初步应用,效果良好。更多还原

【Abstract】 With the development of the society and the advance of the scientific technology, the modern manufacturing enterprises are facing a series of problems, such as unpredictable and fast market competition that is strengthening changeably and constantly. Just under this kind of background, reconfigurable manufacturing system (RMS) becomes one of the most competitive and potential manufacturing pattern. Reconfigurable assembly line (RAL) is a new reconfigurable manufacturing system, which is designed at the outset of rapid change in structure in response to multi-type, varied-toting, multi-function, quick delivery, and sudden changes in market. As an advanced manufacturing system, RAL has become the research focal point of assembly field in recent years. As the key research content of RMS, the design and management of the RAL is an important way that can lead the RMS to further research and application. The reconfigurable assembly line management system can offer a good control and management platform for the enterprises to realize modeling, balancing, scheduling and appraising of the assembly line, to strengthen the flexibility and adaptability of enterprises and to improve the competitive capacity of the enterprises. Hence the research of key technologies for RAL management is of great importance in both theory and application. This paper focuses mainly on the architecture and management technologies of RAL. The main research contents in this paper are as follows:Firstly, the overall framework for RAL is studied. After the system requirement is analyzed for the RA1, the system framework is presented based on the multi-agent technology. The main characteristics of the system structure are analyzed, and the function model of system is designed. The key technologies of system realization are discussed and thus the research direction of the dissertation is determined.Secondly, the modeling technology of RAL is studied. Considering the characteristics of complexity, reconstructing, intelligent and autonomy of the RAL, an agent oriented timed colored Petri net (ATCPN) modeling approach is proposed. Based on the classification of resources, the resource agent is encapsulated so that timed colored Petri net base resource agent is built. The ATCPN model of reconfigurable assembly cell is composed of different resources agent by transition. According to the logical relation of each reconfigurable assembly cell, the transitive relation of message for assembly system is determined and the assembly line could be quickly reconfigured using this method. The performance indexes of ATCPN model for RAL are analyzed, and by studying the transforming from AUML model to ATCPN-based model the ATCPN-based interactive protocol simplification model is presented for the RAL to solve the conflict of the model.Thirdly, the balancing for RAL is studied. The characteristics of mixed-model assembly line for mass customization are analyzed and the special requirements of balancing are studied. The mathematical model is proposed based on three factors, which are integrated including the workstation number, the workstation load and the assembly line efficiency. Then a new hybrid genetic algorithm is developed for finding optimal solution of the problems. In order to prevent the premature convergence problem and enhance the globe-optimization capability, GA (genetic algorithms) is combined with SA (simulated annealing algorithms), and the adaptive crossover and mutation probabilities method are proposed.Fourthly, the scheduling of RAL is studied. A multi-objective optimization mathematical model is proposed, which includes three practically important objectives. Such as minimizing total idle and over workload, keeping constant consuming rate of component and minimizing reconfigurable setup cost are considered. A scheduling method for reconfigurable assembly line is presented based on Pareto multi-objective genetic algorithm, and then chromosome encoding and decoding representation of solution as well as calculations of fitness functions are described. In order to ensure the group’s variety, prevent the premature convergence problem and enhance the globe-optimization capability, some key technologies such as population ranking technique, Niche technique, Pareto solution set filter and elitist selection are applied. The adaptive crossover and mutation probabilities methods are developed.Fifthly, multi-objective scheduling decision-making for RAL is studied. A new nonlinear-objective planning model of combination determining weight is presented based on improved AHP method and information entropy. In view of the grey information in the optimal selection of reconfigurable assembly line multi-objective scheduling, a decision-making model based on fuzzy grey relational analysis is proposed, and evaluation criterion of decision-making is constructed by the close degree of grey relation between decision vector and ideal reference vector which can make the results reliable and credible.Sixthly, the evaluating system for RAL is studied. On the basis of the analysing of the criterion of the evaluation index, a comprehensive evaluation model for RAL is proposed based on the hierarchy analysis according to the characteristics of the RAL and the engineering application, and the meaning of every index is analyzed. The new combination weighting method is used to calculate the weights of every index, and the index is assessed by AHP and fuzzy synthetically evaluation. The subordinative degree of evaluation object relative to optimum reference value adopting least squares principle is regarded as the evaluation criteria.Seventh, the integrated framework of the RAL management system based on CORBA is presented, and the information model for assembly resources is researched. To build and analyze the RAL modeling, the visual modeling and simulation tool for ATCPN is developed.Finally, on the basis of above research, the management system prototype of RAL is developed. The system is applied in the engineering, and the result shows the effectiveness of the system.更多还原