【作者】 王炳刚；

【导师】 邵新宇； 饶运清；

【作者基本信息】 华中科技大学 ， 工业工程， 2010， 博士

【摘要】 随着市场竞争进一步的加剧和产品需求个性化的发展,为了能够及时满足用户多样化的需求,同时又不大量增加企业的库存,近年来,越来越多的企业采用了混流加工-装配系统以提高企业生产的柔性和应对市场需求不断变化的能力,从而提升企业的市场竞争力,比如在轿车发动机、汽车、家用空调等生产企业中,普遍都采用了加工-装配式混流生产系统。为了提高混流加工-装配系统的生产效率,可以采用改善生产设施的方法,比如引进效率更高的生产设备等,但是采用这种方法的成本较高,实施时间较长。相比之下,采用优化混流加工-装配系统生产顺序的方法更具经济性和实用性。本文以轿车发动机混流加工-装配系统为研究背景,对该类系统中的四类典型的优化排序问题进行了深入的研究。通过对轿车发动机生产系统的组成以及各条生产线的性质的分析,对本文要研究的问题进行了归类。按照部件加工线调度、混流装配线排序、加工-装配系统集成优化排序,以及加工-装配系统批量和排序集成优化的顺序分别对各个问题的国内外研究现状进行了系统全面的综述,指出了各个领域存在的问题。对于单条部件加工线调度问题,以最小化最大完工时间为优化目标,对带有限中间缓冲区的多级混合流水部件加工线的调度问题进行了研究。为了确定调度方案,并计算最大完工时间,提出了一种由第一工位投产序列,采用基于事件驱动和空闲机器优先规则相结合的方法。采用了基于遗传算法和模拟退火算法的混合算法求解该问题,在该混合算法中,采用启发式方法和随机产生相结合的方式形成初始种群,结合问题本身的特点设计了新的选择、交叉和变异算子。通过遗传算法和模拟退火算法的混合,克服了各个单一算法的不足,平衡了算法广泛性搜索和集中性搜索的能力。针对相同的问题和计算数据,将该算法的优化结果与近年发表的文献中的结果进行了比较,比较结果验证了该算法的有效性和优越性。然后,按照奇瑞第二发动机公司各条部件加工线的实际构成,应用该公司真实的数据,分别对缸体、缸盖、曲轴和凸轮轴加工线的班次加工计划进行了优化调度,对于每条加工线的调度结果均优于奇瑞第二发动机公司目前采用的调度方法的结果。对于单条混流装配线的排序问题,以部件消耗平顺化和最小化最大完工时间为目标,建立了带有限中间缓冲区的混流装配线的两目标优化排序数学模型。结合多目标混流装配优化排序问题的特点,设计了一种多目标遗传算法用于问题的求解。在此算法中,应用了帕累托分级和共享函数的方法用于可行解适应度值的评价,保证了解的分布性和均匀性,同时对种群初始化、选择、交叉、变异算子以及精英保留策略进行了设计。按照奇瑞发动机公司装配线的实际构成和真实的数据,将该算法的优化结果与采用前一章中的混合算法分别对两个单目标进行优化的结果进行了比较,比较结果验证了该多目标遗传算法的可行性和有效性。对于加工-装配系统集成优化排序问题,研究了由一条带有限中间缓冲区的混流装配线和若干条带并行机和有限中间缓存区的部件加工线组成的拉式生产系统的集成优化排序问题,以平顺化混流装配线的部件消耗及最小化装配线和多条加工线总的完工时间成本为优化目标,提出了系统集成优化框架,基于之前对各条加工线调度和的装配线排序问题的研究,考虑产品和部件的库存约束,建立了集成优化数学模型,提出了一种由装配序列产生各条加工线第一工位加工序列的方法,设计了一种新的多目标遗传算法用于求解该问题,在此算法中,采用了可适应的遗传算子和新的适应度评价函数。对多目标优化算法得到的非支配解集,提出了基于满意度函数的评价方法。按照奇瑞发动机公司各条生产线的实际构成和真实的数据,通过与多目标模拟退火算法的结果进行比较,比较结果验证了该多目标遗传算法的可行性和有效性,应用该可适应多目标遗传算法可以获得满意的非支配解集。为了能够充分利用各个调度区间各条生产线的生产能力,同时为了克服采用完全混流排序方法可能造成频繁切换,以致引起错漏装操作的不足,以装配车间三个连续班次生产计划为输入,以最小化加工-装配系统总的正常完工时间成本、超时完工时间成本和库存成本为目标,对该加工-装配系统的批量和排序集成优化问题进行了研究,建立了优化数学模型,提出了一种基于遗传算法和禁忌搜索算法的混合求解方法,在该算法中,提出了新的编码方式、交叉和变异方法,采用了可适应的交叉和变异算子,应用企业的实际生产数据,将该算法的优化结果与可适应遗传算法的优化结果进行了比较,比较结果验证了该混合算法的有效性。结合奇瑞第二发动机公司的实际需求,对发动机混流生产计划管理的现状、存在问题以及需求进行分析,设计并开发了一套面向发动机混流生产的计划管理软件系统,并把本文提出的优化排序方法应用到该系统中,使研究成果能够在企业中得到实际应用。最后,对全文所做的主要工作进行了总结,并对后续的研究进行了展望。更多还原

【Abstract】 As the development of customization production, more and more mixed-model fabrication-assembly systems are adopted by different manufacturing enterprises, such as in car engine, automobile, air conditioner manufacturing industries, et al., to meet diversified demands of customers without holding large end product inventory. The efficency of these production systems can be improved by two ways. One way is to improve the hardware facilities, for example, by importing more efficient equipments. The shortcomings of this method are that it is expensive and it needs a long period to finish implementation. The other way is to optimize the production sequences. In contrast, the second method is more economic and realistic. In this paper, four typical sequencing problems in car engine mixed-model fabrication-assembly systems are fully addressed.The car engine mixed-model fabrication-assembly systems are analyzed to classify the the typical problems to be considered in this paper, including:scheduling problems in hybrid parts fabrication lines, sequencing problems in mixed-model assembly lines, integrated sequencing problems in mixed-model fabrication-assembly systems and lot sizing and sequencing problems in mixed-model fabrication-assembly systems. And then, the research work in each field is reviewed and the exsiting problems in each field are presented.Then, scheduling problems in hybrid part fabrication flow lines with limited intermediate buffers are considered. The optimization objective is minimizing the makespan. A method, which based on event driven and first available machine rule, is proposed to construct complete schedule and to determine the makespan from the production sequences for the first station. A hybrid algorithm based on genetic algorithm and simulated annealing is proposed to solve the optimization problem, which can balance the algorithm’s exploration and exploitation abilities. In this algorithm, two heuristic approaches and a random generation method are adopted together to form the initial population, new selection, crossover and mutation operators are designed. The feasibility and superiority of the hybrid algorithm is shown by comparing with the methods presented in recently published literature for the same optimization problems. And then, the proposed algorithm is used to solve the real scheduling problems in cylinder body, cylinder cover, crank shaft and camshaft fabrication lines, respectively, in the Second Engine Company of Chery Automobile Co., Ltd. All the optimization results obtained by the hybrid algorithm are better than those obtained by the adopted scheduling method in this company.For sequencing problems in mixed-model assembly lines with limited intermediate buffers, two optimization objectives are considered simultaneously:minimizing the variation in parts consumption and minimizing the makespan. The mathematical models are presented. A multi-objective genetic algorithm are proposed to solve the models. In this algorithm, a new fitness value function is presented based on Pareto ranking and sharing function method to evaluate each individual, which can guarantee the individuals’ diversity and uniformity in the non-dominated solution set. New initialization method, selection, crossover, mutation operators, and elitist strategy are presented. Using the real production data in the Second Engine Company of Chery Automobile Co., Ltd, the multi-objective genetic algorithm is applied to optimize the production sequences for the mixed-model assembly line. The optimization results are compared to the single objective optimization result obtained by the hybrid algorithm proposed in chapter 2 respectively. The comparison results show that multi-objective genetic algorithm proposed in this chapter is an efficient method for sequencing problems in mixed-model assembly lines.Next, integrated sequencing problems in mixed-model fabrication-assembly systems are investigated. The considered production systems are composed of one mixed-model assembly line with limited intermediate buffers and four parts fabrication lines with limited intermediate buffers and identical parallel machines. The optimization objectives include minimizing the variation in parts consumption in the assembly line and minimizing the total makespan cost in the fabrication-assembly system. Based on the research efforts in previous chapters, considering the inventory constraints, the integrated optimization framework and mathematical models are proposed. A three-stage method to determine the production sequences for the first stations of all the fabrication lines from the production sequence for the assembly line is put forward. An adaptive multi-objective genetic algorithm is presented for solving the integrated optimization problem. In this algorithm, adaptive crossover probobility and mutation probobility are adopted to perform genetic operations, and new fitness value function is employed to guarantee the solutions’ diversity and uniformity. A method based on desirability function is proposed for comparing the non-dominated solution sets obtained by multi-objective optimization algorithms. The optimization results of the adaptive multi-objective genetic algorithm is compared to a multi-objective simulated annealing algorithm by using the real production data in the Second Engine Company of Chery Automobile Co., Ltd. The comparision results indicate that the adaptive multi-objective genetic algorithm performs better than the multi-objective simulated annealing algorithm, satisfactory non-dominated solution sets can be obtained by the adaptive multi-objective genetic algorithm.In order to make full use of production capacity in each scheduling period and to avoid too frequent setups and mistake operations caused by complete mixed-model sequencing results, the lot sizing and sequencing integrated optimization problems in mixed-model fabrication-assembly systems are addressed. The lot sizing and sequencing problems consider consequent three shift production plans in the assembly line simultaneously. The optimization objective is minimizing the production cost during normal worktime, production cost during overtime period, and total holding cost during whole planning period. The optimization mathematical models are proposed. A hybrid algorithm based on genetic algorithm and tabu search is presented to solving the mathematical models. In this algorithm, new encoding method, crossover and mutation operators are designed, and adaptive crossover and mutation probobilities are adopted. The superiority of the hybrid algorithm is shown by comparing with an adaptive genetic algorithm using the same real computation data in the Second Engine Company of Chery Automobile Co., Ltd.For practical applications, the existing problems and the demand in the mixed-model production plan management in the Second Engine Company of Chery Automobile Co., Ltd are analyzed. A mixed-model production plan management software system is designed and developed. The sequencing methods proposed in this thesis are employed in the software system and applied to real production plan management.The research work is concluded and the future research efforts are proposed in the last chapter.更多还原