基于分散搜索的多目标混流装配线排序问题研究

发布时间：2018-05-01 13:08

本文选题：分散搜索算法 + 混流装配线　；参考：《华中科技大学》2011年硕士论文

【摘要】：伴随着经济的全球化发展,用户需求越来越趋向于多样化、个性化,制造业已经发展为由传统的大批量生产模式转变为多品种中小批量生产模式。多品种混流装配线正是实现多品种中小批量生产的有效途径。混流装配线排序问题是车间调度的一个分支,是实现混流生产的关键问题之一,对投产顺序的优化可以降低在制品库存,缩短交货期,提高混流装配线运作效率,从而提高企业的竞争力。由于其具有很高的工程应用远景而优化求解又十分复杂,长期以来一直是制造与学术界关心的热点问题。为了解决制造企业中混流装配线投产排序时面临的实际问题,达到缩短装配时间,提高生产效率的目的,本文从工作站生产负荷和基于JIT的时间优化两个方面,提出了包括:最小化工作站的闲置与超载时间、最小化总调整变换时间、交货期满意度、保持各零部件消耗速率均匀优化目标,建立了相应的多目标数学模型。为了满足企业高效、快速的求解多目标模型的需求,本文首先运用了加权权重法对多目标模型进行处理;设计了一种改进分散搜索算法,该算法首次提出了一种基于遗传理论的多样性初始解生成法,不仅保证了初始种群的多样性而且为参考集提供了高质量的初始解。同时,设计了参考集生成更新方法、子集产生方法、子集合并方法和优化解方法等多种机制。加权权重法权重系数的选择缺乏一定的理论基础,所得解不全是非支配解,而多目标进化算法则可以得到一个非支配解集供决策者选择。鉴于此,本文还提出了一种多目标分散搜索算法。基于分散搜索的柔性框架重新设计了部分子方法用于实现对多目标的求解。针对两种多目标处理方法,本文分别对其进行了验证。对改进分散搜索算法,将其与HA、GA等算法的性能进行比较,证明该方法保持了种群的多样性,又寻求到了质量较高的满意解。通过对多目标分散搜索算法与多目标遗传算法的验证结果进行对比,表明多目标分散搜索算法所得的非支配解集能够很好的逼近了Pareto前沿,同时也具有很好的分散性。最后,对全文工作进行了总结并提出了一些未来发展的方向。
[Abstract]:With the development of global economy, the user's demand is more and more diversified and individualized. The manufacturing industry has developed into the traditional mass production mode to transform the multi-variety medium and small batch production mode. Multi-variety mixed-flow assembly line is an effective way to realize multi-variety medium-and-small batch production. The scheduling problem of mixed-flow assembly line is a branch of workshop scheduling and one of the key problems to realize mixed-flow production. The optimization of production sequence can reduce the in-process inventory, shorten the delivery time, and improve the operation efficiency of mixed-flow assembly line. In order to improve the competitiveness of enterprises. Because of its high prospect of engineering application and the complexity of optimization, optimization has long been a hot issue in manufacturing and academic circles. In order to solve the practical problems in the scheduling of mixed-flow assembly lines in manufacturing enterprises, to shorten assembly time and improve production efficiency, this paper focuses on two aspects: workstation production load and time optimization based on JIT. It includes minimizing idle and overload time of workstation, minimizing the total adjustment and transformation time, the satisfaction of delivery time, keeping the consumption rate of parts and components evenly optimized, and establishing the corresponding multi-objective mathematical model. In order to meet the needs of enterprises to solve the multi-objective model efficiently and quickly, this paper first uses the weighted weight method to deal with the multi-objective model, and designs an improved decentralized search algorithm. This algorithm proposes for the first time a genetic theory-based initial solution generation method for diversity, which not only guarantees the diversity of the initial population, but also provides a high quality initial solution for the reference set. At the same time, several mechanisms are designed, such as reference set generation updating method, subset generation method, subset merging method and optimization solution method. The selection of weight coefficients in the weighted weight method lacks a certain theoretical basis, and the resulting solutions are not all non-dominated solutions, while the multi-objective evolutionary algorithm can obtain a set of undominated solutions for decision makers to choose. In view of this, this paper also proposes a multi-objective decentralized search algorithm. The flexible framework based on decentralized search is redesigned to solve multi-objective problem using partial submethod. Two multi-objective processing methods are verified in this paper. The performance of the improved decentralized search algorithm is compared with that of HAGA-algorithm. It is proved that the proposed method maintains the diversity of population and finds a satisfactory solution of high quality. By comparing the results of the multi-objective decentralized search algorithm with the multi-objective genetic algorithm, it is shown that the non-dominated solution set obtained by the multi-objective decentralized search algorithm can approach the Pareto frontier well, and it also has a good dispersion. Finally, the paper summarizes the work and puts forward some directions for future development.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH186

【相似文献】