高精度铜板带配料熔炼过程的智能建模与优化方法研究
文献类型：学位论文
| 作者 | 张浩
|
| 学位类别 | 博士 |
| 答辩日期 | 2014-05-16 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 导师 | 朱云龙 ; 陈瀚宁 |
| 关键词 | 配料优化 熔铸作业调度 生物启发式计算 多目标优化 铜板带 |
| 其他题名 | Research on Intelligent Modeling and Optimization Methods for Burdening and Melting Process of High-Precision Copper Strips |
| 学位专业 | 机械电子工程 |
| 中文摘要 | 高精度铜板带生产是铜加工行业重要的组成部分之一，国家重点支持发展这一系列的产品。高精度铜板带产品具有极高的导电性和导热性，良好的耐腐蚀性和强度高等特性，所以在许多重要领域，尤其是高精尖领域有着广泛的应用。例如，电缆带电缆线等材料主要用于现代通讯产业的发展上；变压器带等用于电力工业上；具有高导电性和高导热性的高强度铜合金可用于电机整流子、电气化铁路架空接触线、电子通讯导电元件、集成电路引线框架和电真空器件等。 高精度铜板带的加工流程长，工艺复杂，如何科学管理生产线对加快生产效率，提高铜板带质量，降低生产成本等有着重要的意义。配料熔铸过程是高精度铜板带生产线的起始工序，为铜板带的生产提供铜铸锭，对后面轧制和精整等工序有着重大的影响，所以对配料熔炼过程的合理优化控制直接关系到整条高精度铜板带生产线的运作。在配料熔炼过程中，配料计算和熔炼过程的作业调度是两个关键部分。本文将着重对这两部分的建模优化进行研究，而优化目标和约束条件多是配料熔炼过程建模优化问题的特点。 生物启发计算是受生物行为启发，模拟生命体功能、特点和作用机理而形成的一类智能优化方法。它在求解大规模计算、NP难等传统方法难以解决的问题方面，表现出了卓越的性能。自从上世纪90年代开始，生物启发计算开始受到了世人的瞩目，大量学者投身于这一领域的研究。目前，很多基于生物启发计算的优秀算法已被应用到各类实际工程问题当中。 实际工程问题中，人们关注的指标往往不止一个，而且一般情况下这些指标都是相互冲突的，所以学者们一直致力于能同时优化多个目标的方法研究。基于生物启发计算的多目标优化方法正处于方兴未艾的的阶段，在多目标策略的设计和生物启发计算方法与多目标策略的结合上都有很多值得研究的地方。因此本论文研究的另一个重点是对已有基于生物启发计算的算法进行改进，使其能够更好地融入多目标策略，从而形成有效的多目标优化算法，并将其应用于高精度铜板配料熔炼过程的模型优化。 论文的主要研究内容包括：高精铜板带配料熔炼过程优化模型研究、多策略的混合多目标人工蜂群算法研究、基于根系生长模型的智能优化算法研究、基于生物启发计算的高精度铜板带配料熔炼过程多目标模型优化研究。具体的研究内容和创新性成果概括如下： (1) 高精铜板带配料熔炼过程优化模型研究 分别针对高精度铜板带配料计算和熔铸过程进行了研究，考虑生产过程中的特点，建立了配料优化模型和熔铸作业调度模型。在建模过程中，考虑到传统意义上的最优解, 难以适应高精铜板带配料和熔炼过程中复杂多变的现场环境，而多种备选的具有多样性的Pareto最优解集配合生产人员的经验更具有可操作性，所以建立多目标模型更为恰当。在配料优化模型中，主要考虑了原料种类、配料原则、熔炼中的烧损和装炉顺序等因素，建立了带有配比、投料顺序和库存等约束，最小化原料成本和最大化带入旧料的两目标配料优化模型。熔铸过程的作业调度模型主要解决的问题是根据铜板带熔铸生产线现有生产能力和熔炼工艺，为实现对客户承诺的交货期，达到降低生产成本的目的，而对生产任务进行调度优化，通过减少熔炼时间和熔炼炉洗炉次数来降低生产成本。熔炼调度模型需要平衡的优化目标有两个，分别是：最小化生产总时间和订单未编入计划而受到的总惩罚值。 (2) 多策略的混合多目标人工蜂群算法研究 首先对单目标优化算法-人工蜂群算法（Artificial Bee Colony, ABC）进行了分析，然后结合人工蜂群算法的特点，融入目前广泛应用的非支配排序、拥挤距离和最近提出的归一化目标值求和等多种多目标策略，提出了混合多目标人工蜂群算法（Hybrid Multi-Objective Artificial Bee Colony, HMOABC）。在此基础上，融入生物演化模式的最高形式-群落进化的思想，提出了多蜂巢多目标人工蜂群算法（Multi-Hive Multi-Objective Artificial Bee Colony, MHMOABC）。以一组标准测试函数为实验环境，对混合多目标人工蜂群算法和多蜂巢多目标人工蜂群算法进行了性能测试，实验结果表明人工蜂群多目标优化算法具有较强的优化能力，能取得非常好的优化结果。同时，与传统的多目标算法NSGAⅡ和MOPSO优化相同函数的结果做了比较，比较结果表明所提多目标算法能够更快速准确的逼近真实的Pareto前沿，优化效果较好。 (3) 基于根系生长模型的智能优化算法研究 生物个体行为在生物启发计算优化理论与模型中处于最底层，是基于群体和群落计算模式的基础。植物根系生长从一粒种子发芽开始，属于生物个体行为。通过对植物根系生长特点研究，提出植物根系生长模型，并利用计算机对植物根系生长行为进行了仿真，证明了植物根系生长模型的正确性。在根系生长模型的基础上，提出了植物根系生长算法（Root Growth Algorithm，RGA）。通过在标准测试函数上的实验，证明该算法不但具有良好优化精度和收敛速度，而且优化效果优于其它经典算法进。结合植物根系生长算法特点，加入多目标优化策略和自适应策略，将根系生长模型实例化为多目标植物根系生长算法（Multi-Objective Root Growth Algorithm，MORGA）用于处理带有约束条件的多目标优化问题。实验结果证明多目标植物根系生长算法对于带有约束条件的多目标优化问题，是一种非常有效的优化技术，并优于其他传统多目标算法。 (4) 基于生物启发计算的高精度铜板带配料熔炼过程多目标模型优化研究 应用混合多目标人工蜂群算法，多蜂巢多目标人工蜂群和自适应多目标植物根系生长算法分别对高精度铜板带配料计算模型和熔炼调度模型进行优化。首先，根据模型和算法特点，设计编码和优化过程，然后提出约束条件处理方法，最后选用基于模糊集合理论的选优方法来对非支配解集进行排序。测试实验是以某铜板带加工企业的真实生产数据为基础，对算法进行测试，并与其它多目标优化算法进行了比较。实验结果表明，混合多目标人工蜂群算法，多蜂巢多目标人工蜂群和自适应多目标植物根系生长算法都能够有效地解决高精度铜板带配料熔炼过程的优化问题，而且效果比其他传统算法更好。 |
| 索取号 | O224/Z36/2014 |
| 英文摘要 | Producing high-precision copper strips is one of the important parts in the copper processing industry, and Chinese government focuses on development of products of high-precision copper strips. High precision copper strips have many excellent properties, including high electrical conductivity, high thermal conductivity, high corrosion resistance and high strength. So high precision copper strips have been widely applied in many important areas, especially, in the highly sophisticated field. For instance, cable and cable line are mainly used for the development of modern communications industry; Transformer belt is used in power industry; copper alloy with high strength, high electrical conductivity and thermal conductivity can be used for motor commutator, overhead contact line of electrified railway, conducting element for electronic communications, IC lead frame and vacuum devices. Due to long machining process and complicated technology of high-precision copper strip, how to manage the production line to increase productivity, improve quality of copper strip, and reduce production costs is significant. Burdening and casting that supply copper ingot for the production of copper strips are the starting part of the production line of high-precision copper strips, which has a major impact on rolling and finishing processes in the back. So reasonable optimization and control for the burdening and smelting processes relate to the operation in the entire production line of high precision copper strips directly. In the burdening and smelting process, burdening calculation and job scheduling of smelting process are the two key parts. This article will focus on the study of modeling and optimization in the two parts. Many optimization objectives and constraints are the features in modeling and optimization of burdening and smelting process. The bio-inspired computing (BIC) is an intelligent optimization method that is inspired by the biological behavior, and simulates the life’s functions, features and mechanism. BIC shows excellent performance when it solves large-scale computing, NP-hard and other difficult problems that are difficult to solve using traditional methods. Since the 1990s, BIC attracted worldwide attention and a large number of scholars engaged in this research field. Currently, many excellent algorithms based on BIC have been applied to all kinds of practical engineering problems. In general, more than one objective is focused on, and in most cases, these objectives are conflicting. So scholars are devoting themselves to research the approaches that can optimize multiple objectives simultaneously. Multi-objective optimization method based on BIC is gathering momentum. It is worthy of study in the design of multi-objective strategies and the combination of BIC methods and multi-objective strategies. So the other research focus in this thesis is how to design effective multi-objective optimization algorithms that can be applied to optimize burdening and smelting process model by means of improvement and integration of the existing algorithms based on BIC and multi-objective strategies. This thesis includes the following contents: the research on the optimization model of burdening and smelting process in the production line of high-precision copper strips, hybrid multi-objective artificial bee colony algorithm using multiple multi-objective strategies, the research on intelligent optimization algorithm based on root growth model, the research on multi-objective optimization based on BIC for burdening and smelting process model in the production line of high-precision copper strips. The details of the research and innovative achievements are summarized as follows: (1) In this chapter, burdening and smelting process in the production line of high-precision copper strips are studied respectively. Considering the characteristics in the process of production, burdening optimization model and job scheduling model of smelting are established. In the modeling process, the optimal solution in the traditional sense is difficult to adapt to complex and changeable environment in the workshop of burdening and smelting process. Because it will be more operable that optionality and diversity of Pareto optimal set are combined with technical staff’s experience, establishing a multi-objective model is much more appropriate than a single objective one. Considering the factors including the category of raw materials, burdening principle, burning loss of raw materials for in the smelting process and charging sequence, the burdening optimization model with the constraints of ratio, feeding sequence and stocks is established with two objectives of minimizing the total cost of raw materials and maximizing the amount of waste material thrown into melting furnace. The problem that job scheduling model of smelting process solve is that in accordance with the existing production capacity and the smelting technology of smelting production line for copper strips, it will schedule and optimize productive task to meet the delivery date of the customer and achieve the purpose of reducing the production cost by means of reducing the smelting time and the number of washing smelting furnace. There are two objectives to be balanced in smelting scheduling model, and respectively, they are: to minimize the total manufacturing time and the total penalty value because of orders which are not being incorporated into the production plan. (2) Firstly, the single-objective optimization algorithm called Artificial Bee Colony (ABC) algorithm is analyzed. Then according to the characteristics of ABC, Hybrid Multi-Objective Artificial Bee Colony (HMOABC) is proposed, which integrates some multi-objective strategies, including non-dominated sorting, crowding distance and summation of normalized objective values that is proposed recently, into ABC. On this basis, Multi-hive Multi-Objective Artificial Bee Colony (MHMOABC) is proposed, which integrates the superlative form in biological evolution - community evolution into ABC. With a set of benchmark functions as the experimental environment, the performance of HMOABC and MHMOABC is tested. The experimental results show that multi-objective artificial bee colony algorithm has excellent optimal performance and can obtain very good optimization results. Then, they are compared with the classic multi-objective algorithm, NSGAⅡ and MOPSO, on these functions. The results show that the proposed multi-objective algorithms can approximate the true Pareto front more fast and accurately than NSGAⅡ and MOPSO. So optimal performance of HMOABC and MHMOABC is better than NSGAⅡ and MOPSO. (3) Biological individual behavior is the first level in the BIC theory and model, which is the basis of calculation mode of swarm and community evolution. Plant roots start to grow from a seed, which belongs to biological individual behavior. Through the study of plant root growth characteristics, plant root growth model is proposed. And the correctness of root growth model is proved by computer simulation of plant root growth. Root Growth Algorithm (RGA) is proposed on the basis of root growth model. Through the experiments on benchmark functions, it’s proved that the algorithm not only has good optimization accuracy and convergence speed, but also its optimization results are better than other classical algorithms. According to the characteristics of RGA, Root Growth model is instantiated as Multi-Objective Root Growth Algorithm (MORGA) for multi-objective optimization problem with constraints when multi-objective optimization strategy and adaptive strategy are integrated into Root Growth model. Experimental results show that MORGA is a very effective optimization method for multi-objective optimization problem with constraints, and is superior to other classic multi-objective algorithms. (4) HMOABC, MOHABC and adaptive MORGA are used to optimize burdening optimization model and job scheduling model of smelting in the production line of high-precision copper strips. First of all, according to the characteristics of the models and these algorithms, the code and optimization process are designed. Then the constraint handling method is proposed. Finally the set of non-dominated solutions is sorted using the selection method based on fuzzy set theory. The testing experiments that used to test the proposed algorithms are based on the real production data from an enterprise processing copper strips, and HMOABC, MOHABC and adaptive MORGA are compared with other multi-objective optimization algorithms in these experiments. Experimental results show that HMOABC, MOHABC and adaptive MORGA can solve the optimization problems of burdening and smelting process in the production line of high-precision copper strips effectively, and are better than other classic algorithms. |
| 语种 | 中文 |
| 产权排序 | 1 |
| 页码 | 132页 |
| 分类号 | O224 |
| 源URL | [http://ir.sia.ac.cn/handle/173321/14802]  |
| 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 | 张浩. 高精度铜板带配料熔炼过程的智能建模与优化方法研究[D]. 中国科学院沈阳自动化研究所. 2014. |
入库方式： OAI收割
来源：沈阳自动化研究所
浏览0
下载0
收藏0
其他版本
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。