【作者】 成浩；

【导师】 胡业发；

【作者基本信息】 武汉理工大学 ， 机械制造及其自动化， 2006， 硕士

【摘要】 现代制造型企业的主要活动之一是生产管理,即利用企业资源,根据生产任务和任务顺序约束制定和执行生产计划。有效的生产调度方法和优化技术的研究和应用是实现先进制造和提高生产效益的基础和关键。本文主要工作分为两个部分: 第一部分为作业车间调度理论算法的研究。具体包括:通过分析对国内外的研究进展,阐述精确算法和近似算法各自的特点。禁忌搜索一种求解作业车间调度问题的有效局部搜索算法,但其搜索最优解的能力取决于初始解的选择。演化算法是通过选择、交换和变异等操作使群体进化来进行全局优化搜索的,它本身不能减小搜索空间的大小,但由于群体搜索的机制使得它能有效地覆盖较大的解空间。将禁忌搜索纳入到演化计算的框架中,增加局部搜索能力,提出禁忌演化算法。新算法通过禁忌策略产生初始种群,并引入了分级策略,对种群按照适应值进行划分,对处于不同级别的个体采用不同的遗传操作。通过对个体的分级,可以区分个体在搜索过程中的职能:优秀的个体进行局部极小值的开采;其它的个体进行搜索空间的探索,以发现新的局部极小值。通过测试Benchmark问题,数值实验表明算法收敛速度快,结果较好。 第二部分为禁忌演化算法的应用研究。具体包括:测试与JSSP类似的组合优化问题TSP问题,并阐述将禁忌演化算法植入实际调度系统,生产排产系统中的开发过程。通过分析智能算法应用于实际调度系统中失败的原因,以此为基础提出一种智能调度框架,该框架的特点在于生产反馈信息的收集和调度器使用多种算法共同决策的模式。更多还原

【Abstract】 Production Management is one of main activities in modern manufacture enterprise, which executes production scheduling by allocating resources in order to perform a number of tasks, which are assigned to resources in a temporal order. The research and application of effective production scheduling methods and optimization techniques are the key elements to implement modern manufacture and promote production efficiency. This paper consists of two major parts.The 1st part is the research on theory algorithms of Job Shop Scheduling Problem (JSSP). By analyzing the domestic and foreign research developments, this paper notes exact and approximate methods’ characteristics respectively. Tabu Search is an effective local search algorithm for the job shop scheduling problem, but the quality of the best solution found depends on the initial solution. Evolutionary algorithm is a global search algorithm by choosing, crossover and mutation operations to operate the population. The method itself cannot decrease the search space, but the population search mechanism can cover a large scale solution spaces. To overcome this problem we present a new Evolutionary-Tabu alogrithm (ETA) that uses a population of Tabu Search runs in an Evolutionary Computation framework. ETA adopts a search strategy to classify the individuals by their fitness. Individuals’ classification differentiate respective function in search process, that’s the excellent individuals mine the local optimal solution and others explore the search domain to find new local optimal solution. By testing benchmark instances the results show the new algorithm is satisfactory.The 2nd part is the application of ETA. Testing Traveling Salesman Problem (TSP), which is a similar combination optimization problem to JSSP. In addition, we explain how to design production scheduling system by planting ETA into a real-world scheduling system. Based on analyzing the failures using intelligent algorithms in real-world scheduling system, we put forward an intelligent scheduling framework. The characteristics of the framework are collecting the production response information and the judge modal of scheduler using multi algorithms.更多还原