【作者】 王秀萍；

【导师】 杨立熙；

【作者基本信息】 福州大学 ， 管理科学与工程， 2018， 硕士

【摘要】 柔性制造和绿色制造是《中国制造2025》规划中的主线,也是制造企业转型升级的关键。由于多目标柔性作业车间调度问题最契合柔性化生产的实际环境,逐渐成为车间调度领域研究的重点,因此研究多目标柔性作业车间低碳调度问题具有重要理论与实践意义。本文混合免疫遗传算法,研究多目标柔性作业车间低碳调度问题的建模与求解。首先,基于对机械加工过程的分析,建立了结合空载功率和加工时间的机器能源消耗计算公式,进而在传统柔性作业车间调度模型的基础上增加新的约束条件,并从生产效率、设备利用率、能源资源三个方面建立以最大完工时间最小、瓶颈机器负荷最小、能源消耗总量最小为目标的新型调度模型。其次,为求解调度模型,本文结合免疫算法和遗传算法提出基于记忆指导的免疫遗传算法(MG-IGXA)。算法包括遗传算法模块(GA)、记忆库模块(ML)和免疫算法模块(IA)三部分,其中以GA为基本框架,引入IA和ML弥补GA存在的局部收敛、缺乏对优良个体记忆等方面的不足,以提升算法效率。在GA模块中,结合基于工件剩余加工时间最大的启发式规则和随机化法方法初始种群,改善初始解质量;采用多元化的遗传操作寻优,在交叉操作中利用相似度阈值引入ML模块的精英解,指导种群优化的方向。ML模块使用记忆库保存每次迭代产生的精英解,避免丢失精英解从而导致算法退化。在IA模块中,基于海明距离计算个体的亲和度、相似度阈值和个体浓度,根据相似度阈值消除冗余个体,防止后代生成冗余信息;采用与初始种群相同的方法生成新个体,并由相似度阈值决定其是否加入种群,以保证种群多样性。针对多个优化目标,对种群进行修剪保证算法收敛至均匀分布的Pareto曲面。最后,在基准算例的基础上添加与能耗相关的功率数据以测试MG-IGXA。以解的质量和分布度作为MG-IGXA和MOGA的评价标准,对26组基准算例进行测试,每组算例都连续运行20次,共520组实验结果。结果表明,MG-IGXA能得到质量更优、分布更均匀的Pareto非支配解集,而且随着问题的增加,MG-IGXA的优化程度更为显著,进一步验证了模型和算法的可行性和优越性。更多还原

【Abstract】 Flexible manufacturing and green manufacturing are the main line in the planning of made in China 2025,and also the key to the transformation and upgrading of manufacturing enterprises.Because the multi-objective flexible job shop scheduling problem is the most compatible with the flexible production environment,it has gradually become the focus of the researches on the shop floor scheduling,so it is of great theoretical and practical significance to study the multi-obj ective flexible j ob shop low carbon scheduling problem.In this paper,a hybrid immune genetic algorithm is used to study the modeling and solution of low carbon scheduling problem in multi－objective flexible job shop.First,based on the analysis of the machining process,a formula for calculating the energy consumption of the machine is established,which combines the unloaded power and the processing time.Then on the basis of the traditional flexible job-shop scheduling model,new constraints are added,and from three aspects of production efficiency,equipment utilization,energy resources to establish a new type of scheduling model with the minimum makespan,the bottleneck machine load the smallest,the target of minimizing the total energy consumption.Second,the immune genetic algorithm(MG-IGXA)based on memory guidance is proposed to solve the scheduling model by combining the immune algorithm and genetic algorithm.The immune algorithm is suitable for comprehensive coarse grain search,but the search precision is poor.In the genetic algorithm,the genetic manipulation of chromosomes can be carefully searched,but it is easy to get into local optimum.Combining the two algorithms can realize complementary advantages,and the immune algorithm can make up for the shortage of local convergence of the genetic algorithm,and the genetic algorithm can improve the search precision of the immune algorithm.MG-IGXA includes genetic algorithm module(GA),memory library module(ML)and immune algorithm module(IA).GA module,combined with the maximum heuristic rule based on work piece residual processing time and the initial population of random method,improved the initial solution quality;A plurality of genetic manipulations are used to optimize the population.In the cross operation,the similarity threshold is used to introduce the elite solution of the memory bank and guide the direction of population optimization.The ML module uses the memory library to save the elite solution generated by each iteration,so as to avoid losing the elite solution and leading to the degradation of the algorithm.In the IA module,the affinity,similarity threshold and individual concentration are calculated based on the distance of the hamming distance,and redundant individuals are eliminated according to the similarity threshold to prevent the generation of redundant information from generation.The new antibody was generated in the same way as the initial population,and the similarity determines whether the new antibody is added to the population to ensure the diversity of the population.The population was then trimmed and distributed evenly.In the end,the power is divided into two classes,between 0-1 and 1-5,to avoid the impact of the numerical disparity of power.the paper uses 26 benchmark examples of ntwo kinds of power to test the quality and distribution of the solution.Each group of examples runs 20 times in a row,received 520 groups experiments.The experiment shows that the proposed algorithm can obtain Pareto solution with better quality and more uniform distribution,which fully verifies the feasibility and effectiveness of the model and algorithm.更多还原