【作者】 胡凯林；

【导师】 李平； 宋春跃；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2011， 博士

【摘要】 钢铁工业是我国国民经济的重要支柱产业,其在快速发展的同时也面临着高成本、高能耗等诸多问题和挑战。以散装原料场为核心的原料贮运与处理工序,作为钢铁生产的第一步骤和钢铁企业供应链的重要环节,在降低原料成本、节能降耗方面扮演着重要角色。但一直以来,对钢铁企业散装原料场运行优化相关的研究都没有得到应有的重视,其运行管理的信息化与智能化在整个产业中也处于较低水平。因此,对钢铁企业散装原料场运行过程中的调度和优化问题进行研究,提高其生产效率和管理质量,不但可以填补相关研究领域的空白,对解决我国钢铁工业面临的问题也有着重要的现实意义。本文在综述相关组合优化问题基本理论和研究现状、介绍分析钢铁工业散装原料场生产运行流程的基础上,对原料场运行过程中的几个主要问题进行抽象、定义,将其转化为特定类型的组合优化问题,建立数学模型,并设计了相应的求解算法。并且,通过将以上模型、算法与先进的定位与监控技术相结合,设计了一种钢铁企业原料场运行监控与优化系统,并在实际企业实现了应用。本文的主要工作和创新点如下：(1)提出了一个描述钢铁企业散装原料场在有限生产周期内的存储空间分配问题的优化模型。该模型通过将原料出/入库过程离散化、对场地空间和原料料堆进行分块,以及建立一个复现原料场调度人员决策过程的双层决策框架,将原料场存储空间分配过程转化为一个多目标组合优化问题。作为问题的优化目标,该模型还对散装原料场混料风险以及场地空间利用率的量化指标进行了定义。(2)提出了一种适用于求解原料场存储空间分配问题的Pareto多目标蚁群优化算法DCACO。该算法中,对应于问题目标函数的多个种群使用各自独立的信息素更新策略,在一个分散/集中框架引导下进行寻优,有效地改善了所求解集的收敛程度和在目标空间的分布情况；另外,该算法采用多信息素轨迹矩阵构架,独立记录每个生产周期内每个种群的信息素轨迹,使搜索过程有更好的指向性,提高了寻优效率。(3)提出了一个描述面向炼铁原料混匀过程中原料场堆取料机调度问题的离散调度模型。该模型通过将原料场中的物料流抽象为“零件”和“产品”,并对原料场场地以及堆取料机移动路径进行划分,使炼铁原料混匀过程中的各种事件同步于一个离散时间轴上,从而将该过程转化成为一个带有有限中间存储、顺序相关准备时间的二阶组装流水车间调度问题(F2|STsd,block,assembly| Cmax)。在此基础上,考虑多堆取料机协同操作的情况,并在模型中加入符合实际工业过程的约束,将该问题拓展为一个带有有限中间存储、顺序相关准备时间、组装过程以及机器资格约束的二阶混合流水车间调度问题(HF2|STsd,Mj,block, assembly|Cmax)。(4)提出了两种适用于求解单机条件下的炼铁原料混匀过程堆取料机调度问题的元启发式方法——包含种子染色体和“批次合并”算子的改进遗传算法iGA,以及带有禁忌定时器的改进蚁群算法ACOtt。这两种算法分别从各自角度,在传统遗传算法/蚁群算法基础上加入基于问题特性和知识的局部搜索/启发式过程,有效提高了其在求解该种类型调度问题时的性能。(5)提出了一种适用于求解多机条件下的炼铁原料混匀过程堆取料机调度问题的蚁群算法ACO_AHFLB。该算法将蚁群中单个蚂蚁的寻优路径分解为多条子路径,分别用于问题中每台并行机上生产序列寻优,并使用多个信息素轨迹矩阵,对每条子路径上的信息素轨迹进行单独记录,提高了搜索效率；同时设计了一个子路径切换机制,使各条子路径在寻优过程中的时间轴尽可能保持同步,从而减少死锁现象的发生,提高了求解质量。(6)设计了一个钢铁企业原料场实时监控与运行优化系统BMYRMOOS,并在实际钢铁企业实现了应用。该系统通过将本文所研究的原料场运行调度与优化问题的数学模型及其求解算法与差分定位、三维显示等技术相结合,实现了对钢铁企业散装原料场储空间分配情况和堆取料机工作状态的动态监控,并且根据料场生产计划与运行状态,为管理人员提供存储空间分配和堆取料机调度的优化决策方案,有效提高了钢铁企业原料场的管理质量和运行效率。更多还原

【Abstract】 Iron and steel industry is an important pillar industry of the national economy. Along with the rapid development, it is also facing issues such as high material cost and high energy consumption. As the first step of the iron and steel production line, the bulk material yard which is responsible for the stroage and processing of raw material, plays an important role in cost reduction and energy conservation. However, for a long time, researches that are about the scheduling and optimization problems in bulk material yard are rarely found in the literatures. On the other hand, comparing to other procedure in the iron and steel production, the bulk material yard operation lacks for effective means of optimization. Hence, research on the operation scheduling and optimizaiton problems in bulk material yard of iron and steel industry not only expands the research area of the industrial operations, but also has important practical significance.Based on the a review of the related researches on the combinatorial optimization problems and an introcution of the operation process of the bulk material yard, this thesis focus on some main scheduling and optimizaiton problems in the bulk materal yard operation and converts them into different types of combinatorial optimization problem. The corresponding mathimatical models and algorithms are also proposed. And by combining these models and algorithms with the technologies of positioning and monitoring, an real-time monitoring and operation optimization for bulk marterial yard is also designed and implemented in a real iron and steel plant. The main contributions of this thesis are presented as follows:(1) A model for the optimization of bulk material yard storage space allocation in a finite horizon is proposed. In the model, the process in real industry is convert into a multi-objective combinatorial optimization problem via discretizing the inbound/outbound operations and partitioning the yard space and material piles, while a 2-layer decision making framework is also established to recreate the decision making process of the materail yard dispathcer. Moreover, the quantified measures for the material blending risk and the space utilization of the yard, which are the two main optimization objectives of the storage space allocation, are also defined in the model.(2) A Pareto multi-objective ant colony optimization algorithm for the storage space allocation optimization, DCACO, is proposed. In the algorithm, two separated colonies are generated to search for each single objective, under indepent pheromone update rules and a distributed/centralized framework, which improves the convergence and the spread extent of the solution set. The multi-pheromone trail matrix strategy is also implemented in the algorithm, which gives the colonies more accurate guidance for search, and improves the computational efficiency.(3) A model that describes the bucket-wheel excavator scheduling in the bulk ore blending process of iron making is proposed. By discretizing the time horizon and partitioning the yard space, eventually the bulk ore blending process is reformulated as a two stage assembly flow shop with sequence-depended setup time and limited intermediate buffer (F2|STsd, block, assembly| Cmax). Further more, the senario of muliple excavator cooperation and some other constraints in real industry are also considerd. And the model is generalized as a two stage hybrid flow shop with sequence-depended setup time, limited intermediate buffer, eligibility constraint and assembly process (HF2| STsd, Mj, block, assembly| Cmax).(4) An improved genetic algorihm iGA and an improved ant colony optimization algorithm ACOtt for solving the single-excavator scheduling in bulk ore blending process are proposed,. Either algorithm is combined with problem knowledge-based local search procedure or heuristic, which improves the both the search speed and the optimal solutions’quality.(5) An ant colony optimization algorithm ACO_AHFLB for solving the multi-excavator scheduling in bulk ore blending process is proposed. In the algorithm, the searching path of each ant is consist of mutiple sub-pathes, for searching the optimal schedulings on the parallel machines. Multiple pheromone matrices are used to improve the search effeciency. And a sub-path switch mechanism is also designed to synchronize the time horizons of all the sub-paths, in order to reduce the probability of deadlock during the solution construction.(6) A real-time monitoring & operation optimization system for bulk material yard (BMYRMOOS) is designed and implentmented in a real iron and steel plant. By combining the mathmatical models and algorithms proposed in this thesis and technologies such as the differential global positioning and 3-D graphic display, the system is capable to realize the real-time monitoring of the storage space allocation and the excavators’status, and provide optimal decision support for the operations management on the bulk material yard of iron and steel industry.更多还原