QQ客服 官方微博 反馈留言

题名	仿生智能算法研究及其在铝加工排产中的应用
作者	亓祥波
学位类别	博士
答辩日期	2016-11-21
授予单位	中国科学院沈阳自动化研究所
导师	朱云龙
关键词	优化算法 仿生计算 群体智能 作业排产 置换流水车间调度 铝加工
其他题名	Research on Bionic Intelligent Algorithm and its Applicaton in Aluminum Production Scheduling
学位专业	机械电子工程
中文摘要	本课题主要以面向高端铝材的铝加工生产为背景，着重对铝挤压的作业排产进行建模研究、对求解排产问题的仿生智能优化方法进行了深入研究。论文的主要研究内容包括：仿植物根系生长行为的优化算法研究、基于精英策略的人工蜂群算法研究、基于双种群协同学习算法的置换流水车间调度优化研究、铝挤压作业排产优化研究。具体的研究内容和创新性成果概括如下：(1)仿植物根系生长行为的优化算法研究。鉴于植物根系在生长过程中表现出的智能决策行为，对植物根系生长特点进行了研究，建立了基于根系生长行为的计算模型。通过对植物根系生长模型的实例化，提出了一种人工根群优化算法。在标准测试函数以及聚类问题上的实验表明了提出的算法具有良好的优化精度和收敛速度。人工根群优化算法为新型生物启发式算法的研究提供了参考。(2)基于精英策略的人工蜂群算法研究。对人工蜂群算法进行了分析，然后结合人工蜂群算法的特点，提出了基于精英策略的人工蜂群算法。首先引入单个精英寻优策略到蜂群算法，设计了求解连续问题的可学习蜂群算法，该算法有效地克服了蜂群算法不能利用算法中已经找到优秀基因的缺陷。其次，提出了一种基于双种群协同学习的算法，按照个体适应度排序后将种群划分为两个种群，分别为每个种群设置了生成新解的策略，高适应度种群自学习，低适应度种群向高适应度种群学习，实现了高适应度种群引导优化的“精英”指导策略。实验验证了所提算法在高维连续问题上具有较强的寻优能力。(3)基于双种群协同学习算法的置换流水车间调度优化研究。为了利用仿生智能算法求解置换流水车间调度问题（PFSP），沿用双种群协同学习算法的实数编码规则，提出了基于SPV与LOV的双种群协同学习算法，并采用基于NEH方法对解初始化和局部搜索，提高了解的质量。以最小化最大完工时间为求解目标在Reeves实例与Taillard实例上进行实验，结果表明基于SPV与LOV的双种群协同学习算法在解决PFSP问题的有效性。(4)铝挤压作业排产优化研究。针对铝挤压的生产特点，将铝挤压排产问题抽象为基于批量组织生产的流水车间调度问题，建立了基于批量处理的PFSP模型。根据铝挤压排产问题组合排列特性，构造了基于直接编码与变邻域搜索策略的双种群协同学习算法。依据实际的工艺路线采用提出的算法求解铝挤压排产问题，结果表明了所建模型的可行性和所提算法的有效性。
英文摘要	With high-end aluminum production line as the background, this subject modeled the key problem of aluminum extrusion production scheduling and studied intelligent optimization methods fit for it. The main contents of the thesis include the following aspects: the research of optimization algorithm simulating plant root growth behavior, artificial bee colony algorithm based on the elitist strategy, a double population co-learning algorithm for permutation flow-shop scheduling problems and aluminum-extrusion-oriented job scheduling and optimization. The details of the research and innovative achievements are summarized as follows: 1) In view of the intelligent behavior of the plant root system in the process of root, plant root growth characteristics were studied and a computational model based on root growth is built. By instantiating the plant root system growth model, an artificial root mass optimization algorithm is proposed. The experiments on standard test functions and clustering problem demonstrate the proposed algorithm has a good optimization precision and convergence speed. The artificial root mass optimization algorithm provides a reference for studying new intelligent optimization algorithms. 2) Based on Artificial Bee Colony (ABC) algorithm, a learnable artificial bee colony (LABC) algorithm is proposed. The LABC algorithm effectively overcomes the defect that ABC can’t use the excellent genetic found in constant iteration. Then a double population co-learning (DPCL) algorithm is proposed. A population is divided into two populations according to their fitness. The individuals of each population are updated according to the given learning rules. The population with high fitness is responsible for exploiting new solution by self-learning. The population with low fitness is responsible for exploring new solution by learning from the population with high fitness. The algorithm realizes elite guidance strategy. The experiment proved that the proposed algorithms have a strong optimization ability of solving high dimension problems. (3) In order to solve permutation flow-shop scheduling problems (PFSP) using the DPCL algorithm, according to the real number coding rules of DPCL, a double population co-learning algorithm based on the Smallest Position Value (SPV) rule and the Largest Order Value (LOV) rule is proposed. The algorithm employs NEH method to initialize a solution and uses a NEH-based method to be a local search strategy. With the objective to minimize the maximum completion time, some simulations were done. Reeves instances and Taillard instances were used in the simulations. The results demonstrate the effectiveness of the DPCL algorithm for PFSP. (4) On the basis of studies of Aluminum extrusion production characteristics, an Aluminum extrusion scheduling model based on a PFSP model of batch processing was built. According to the combination features of Aluminum extrusion production scheduling, a discrete version double population co-learning algorithm employing direct coding and variable neighborhood search strategy is proposed. The proposed algorithm is used to solve the model on the basis of the actual process. The simulation experiment results demonstrate the feasibility of the model and he effectiveness of proposed algorithm.
语种	中文
产权排序	1
页码	115页
内容类型	学位论文
源URL	[http://ir.sia.cn/handle/173321/19454]
专题	沈阳自动化研究所_信息服务与智能控制技术研究室
推荐引用方式 GB/T 7714	亓祥波. 仿生智能算法研究及其在铝加工排产中的应用[D]. 中国科学院沈阳自动化研究所. 2016.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。