【作者】 刘海龙；

【导师】 刘继印； 赵国栋； 赵任；

【作者基本信息】 东北大学 ， 控制工程， 2022， 硕士

【摘要】 在对热轧生产进行调度时,需要考虑诸多因素,进行高效且合理的板坯物流操作,将影响热轧工序的生产效率。热轧作为钢铁生产的中间工序,制定合理有效的热轧生产调度优化方案,将直接影响热轧工序的生产效率,提高物流生产平衡,进而提高钢铁企业的生产效率。本文针对某钢厂热轧产线实际生产需求,提出考虑板坯物流操作的热轧生产调度优化问题,建立热轧生产调度多目标数学模型,设计改进多目标差分进化算法(Multi-Objective Optimization Differential Evolution,MODE),并结合数据解析技术改进算法,开发热轧生产调度工业软件及可视仿真。本文的主要研究内容为:1)针对热轧生产调度优化问题,建立多目标优化模型。以最小化热轧计划中相邻板坯之间的过渡惩罚、最小化板坯库空间倒垛次数、最小化热轧连铸时间差为目标,考虑热轧生产中的工艺要求、板坯库物流操作、连铸热轧生产时间等约束。2)针对所建立的数学模型中目标函数之间存在一定的相关性、生产约束较多、实际问题规模较大等问题,设计了改进的自适应MODE算法。首先通过对个体进行编码,并以随机方式及启发式算法的解作为种群的初始化,并提出了变异策略与控制参数的组合自适应策略,及带有分解思想的选择策略的改进思路。最后基于实际生产数据进行实验,与基于非支配排序的遗传算法(Non-Dominated Sorted Genetic Algorithm-Ⅱ,NSGA-Ⅱ)、标准MODE算法进行对比实验,通过三种指标比对,验证了改进的MODE算法的有效性。3)根据热轧工序实际生产的需求,为了提升算法运算效率,对改进MODE算法进行进一步改进,提出了基于K近邻域(K-Nearest Neighbor,KNN)的代理模型方式改进MODE算法(K-MODE),通过KD(K-Dimensional)树实现KNN算法,在个体适应度进行比较时,用训练后的KNN模型对个体目标函数值进行预测。并在相同时间下将K-MODE与改进MODE进行比较实验,得到良好的改进效果。4)开发热轧生产调度优化工业软件及可视仿真,该工业软件以某钢铁厂热轧的实际需求出发,实现对板坯数据进行解析、调整,并利用MODE算法对热轧生产计划进行编制,及实现动态调整与参数维护功能。并对热轧工序的板坯库进行可视化建模,实现板坯的出库、入库、倒垛操作的可视仿真。更多还原

【Abstract】 When scheduling hot rolling production,many factors need to be considered.Efficient and reasonable slab logistics operation will affect the production efficiency of hot rolling process.Hot rolling is an intermediate process of steel production.Formulating a reasonable and effective hot rolling production scheduling optimization scheme will directly affect the production efficiency of hot rolling process,improve the logistics production balance,and then improve the production efficiency of steel enterprises.According to the actual production demand of a hot rolling production line in a steel plant,this paper puts forward the hot rolling production scheduling optimization problem considering slab logistics operation,establishes the multi-objective mathematical model of hot rolling production scheduling,designs the improved multi-objective optimization differential evolution,improves the algorithm combined with data analysis technology,and develops the hot rolling scheduling optimization system.The main research contents of this question are:1)Aiming at the hot rolling production scheduling optimization problem,a multiobjective optimization model is established.In order to minimize the transition penalty between adjacent slabs in the hot rolling plan,minimize the number of stacking in the slab warehouse space and minimize the time difference between hot rolling and continuous casting,the constraints such as process requirements in hot rolling production,slab warehouse logistics operation,continuous casting and hot rolling production time are considered.2)Aiming at the problems of certain correlation between objective functions,more production constraints and large scale of practical problems in the established mathematical model,an improved adaptive mode algorithm is designed.Firstly,the individual is coded,and the solution of random method and heuristic algorithm is used as the initialization of the population.The combined adaptive strategy of mutation strategy and control parameters and the improvement idea of selection strategy with decomposition idea are proposed.Finally,experiments are carried out based on the actual production data,and compared with the Non dominated Sorted Genetic Algorithm Ⅱ(NSGA-Ⅱ)and the standard mode algorithm.The effectiveness of the mode algorithm is verified through the comparison of three indexes.3)According to the actual production requirements of hot rolling process,in order to improve the operation efficiency of the algorithm,the improved mode algorithm is further improved.An agent model based on K-nearest neighbor(KNN)is proposed.The improved mode algorithm(K-MODE)realizes the KNN algorithm through KD tree.When comparing individual fitness,the trained KNN model is used to predict the value of individual objective function.At the same time,k-mode is compared with the improved mode,and a good improvement effect is obtained.4)The hot rolling production scheduling optimization industrial software and visual simulation are developed.Based on the actual demand of hot rolling in an iron and steel plant,the industrial software can analyze and adjust the slab data,compile the hot rolling production plan by using the MODE algorithm,and realize the functions of dynamic adjustment and parameter maintenance.The visual modeling of slab warehouse in hot rolling process is carried out to realize the visual simulation of slab delivery,storage and stacking.更多还原