【作者】 王志刚；

【导师】 王伟；

【作者基本信息】 大连理工大学 ， 控制理论与控制工程， 2010， 博士

【摘要】 冷轧薄板是钢铁企业中具有高附加值的深加工产品。生产计划与调度水平的提高是冷轧薄板企业增加效益、降低成本、提高竞争力的重要途径。本文依托国家“863计划”课题,以上海宝钢冷轧薄板厂为研究背景,针对该厂冷轧生产线计划编制与作业调度中存在的具体问题,开展了深入系统的调研,完成了以下主要工作：罩式炉退火是冷轧生产线中工艺与其他工序相差较大的“瓶颈”工序,伴随着复杂的资源约束、能源约束和工艺约束,难于解析建模和优化调度。罩式炉退火生产调度包括装炉组合和作业调度两部分。本文首先研究了钢卷装炉组合优化问题,建立了以最小化钢卷组炉的总加热时间为优化目标的数学模型,综合考虑了影响钢卷组合堆垛加热时间的各种因素。提出了一种改进的分组自适应遗传算法,通过工艺规则的指导改善初始解,对遗传基因启发式交叉和变异,同时结合局部搜索的方法实现了对模型的优化计算。罩式炉退火工艺的特殊性决定了它是生产监控的盲区,其生产流程的准确模拟是冷轧多机组排产的基础。本文在装炉组合优化的基础上,进一步研究了罩式炉退火车间调度建模问题。基于大规模集成电路硬件与罩式炉退火车间过程调度之间在并发性、时序性和层次性等基本特点上的相似性,将退火车间工件的流动类比映射为硬件建模问题。采用电子设计领域成熟的SystemC仿真平台以硬件描述的方法建立了罩式退火炉车间调度的仿真模型,能够准确、高效地实现退火工序的流程模拟。在上述工作基础上研究了冷轧生产多品种、小批量情况下的合同交货期优化调度问题。考虑交货期满意度和实际调度中存在的不确定性,采用一种具有模糊处理时间的JobShop调度模型来描述冷轧生产线的物流情况。考虑存在原料前库作为生产缓冲环节的情况,分类定义了合同批量加工时间推进方法的计算规则,以减小模型累计递推误差。运用一种多子种群并行离散粒子群算法对该模型进行优化求解,提高了运算速度。采用Job Shop模型对冷轧多机组生产描述离实际仍有一定距离。本文建立了一个实现冷轧多机组合同生产计划和作业调度的整体优化模型。模型以最小化各机组合同延期惩罚和生产类型切换虚拟成本为优化目标。针对模型整体求解困难的情况,以时间窗推理方式对存在的生产相关情况进行解耦,将问题转化为时间窗约束条件下的单机组计划调度优化问题,以工艺规则为启发采用契合问题本质的改进蚁群算法对其进行求解。现场运行情况表明以本文工作为核心的冷轧生产调度系统是有效可行的,为企业生产计划编制提供了合理指导。更多还原

【Abstract】 The cold-rolled sheet is a kind of deep processed product with high added value in iron and steel enterprise. Improvement of production planning and scheduling level is an important way to increase economic benefits, reduce production costs and enhance competitive power. Based on a key project of National High-Tech Research and Development Program, this dissertation studies the production planning and scheduling problem of cold rolling line in a cold-rolled sheet plant of Shanghai Baosteel Co. Ltd. Following research work is mainly carried on in this dissertation.As the "bottle-neck" process, bell-type batch annealing is obviously different from other processes in the cold rolling line. Its production has complicated resource, energy and process constraints. Its mathematic model is difficult to formulate and optimal scheduling is hard to implement. Scheduling in batch annealing shop includes combinatorial stacking of coils and order optimization. A mathematical model is set up to minimize the total heating treatment time of batch stacking. The factors of affecting heating treatment time are thoroughly included in this model through analysis of technical rules. An improved group adaptive genetic algorithm is proposed to optimize the model with the methods of improving initial solutions, heuristic crossover, mutation and local search under the guidance of process rules.The unique process rules of bell-type annealing lead to its absence from production supervision. Accurate simulation of its process is the base of production planning of the whole cold rolling line. Modeling problem of bell-type annealing shop scheduling is discussed based on combinatorial stacking. The traveling flow of lots in annealing shop is mapped into hardware design modeling analogously, based on the similarity of concurrency, sequence and hierarchy between large-scale integrated circuit and production scheduling of bell-type batch annealing. A simulation model of annealing shop scheduling is developed with hardware description method based on mature SystemC simulation platform in the field of electronics design. Accurate and efficient simulation of batch annealing workflow can be achieved by this model.The optimal order scheduling problem with many categories and small batches is researched based on above work. In order to improve order delivery satisfaction and address scheduling uncertainty, a fuzzy Job Shop scheduling model is established to represent the whole logistics of the cold rolling line, in which an order batch is regarded as a fundamental scheduling job. Due to the inventory in front of each process as the production buffer, manufacturing time progress rules on each process are defined based on different scenarios. A Parallel Discrete Particle Swarm Optimization (PDPSO) algorithm based on multi-sub colony is implemented to solve the fuzzy scheduling model, which improves the calculation speed.Description of order planning for multi-process cold rolling line using Job Shop is still not a suitable way according to its abstraction, so an optimization model for production planning and coil scheduling in the whole process of cold rolling is proposed in this paper. Objective of the model is to minimize order delay punishment and virtual switch cost in each process. It is difficult to optimize the model directly because of its complication, so a reasoning method based on time windows is constructed to deal with the coupling essence of the problem. Then optimization of multi-process production planning is converted to several constrained single-process scheduling problem with time windows. Heuristic improved Ant Colony Optimization (ACO) algorithms are adopted to solve the problem because of resemble between them.The research work mentioned above is embedded in the production management and decision system of the plant as core algorithms. The feasibility and validity of the research work is testified by the running result in the field. It shows that the achievement can provide reasonable guidance for production planning and scheduling in cold rolling line.更多还原