量子进化算法的改进研究及其在轧制规程优化中的实践

发布时间：2018-02-27 08:51

本文关键词： 冷连轧机规程优化量子进化算法多目标优化支持向量机　出处：《燕山大学》2014年博士论文　论文类型：学位论文

【摘要】：冷轧带钢生产在我国国民经济中占有重要的地位。目前，虽然我国钢铁产量巨大，但是整体生产水平偏低，尤其是冷轧板带生产的核心技术多掌握在国外公司手中。因此，，建造拥有自主知识产权的现代化冷连轧生产线，生产出高质量的冷轧带钢，增强产品在市场上的竞争力是国内科研人员的共同目标。轧制规程设定是冷连轧工艺的核心技术之一，建立更合理的轧制规程是提高冷轧带钢质量的有效途径之一，在这方面我国与国际先进水平尚有一定的差距，必须进行更全面、更深入的研究，才能赶超国际先进水平。本文结合某厂1450mm全连续五机架冷连轧机工艺优化计算机系统研发的工程实践，对冷连轧机工艺优化计算机系统进行了基于量子进化算法和支持向量机的轧制规程优化研究与实践。轧制工艺数学模型是进行轧制规程优化的基础，针对不同的工艺优化目标可以建立不同的目标函数及约束条件。本文对量子旋转门角度更新策略进行了研究，通过改进量子旋转门角度更新查询表提升了量子进化算法的收敛性能；以能耗最低为目标，应用改进的量子进化算法对轧制规程进行优化，使轧制总功率降低3%以上。将其应用于某厂五机架冷连轧机组的规程计算中，自2011年3月投产以来，节约了大量的能源，创造了巨大的经济效益。在传统量子进化算法中，应用查询表对量子旋转门角度进行更新时必须针对具体问题具体设计查询表，通用性较差。为了克服这一缺点，引入粒子群算法和微分进化算法，通过其在角度空间进行启发式搜索的方式进行量子旋转门角度更新，提出了两种混合量子进化算法。通过标准函数测试表明，混合量子进化算法增强了量子进化算法的全局收敛性能，提高了通用性，将其应用于轧制规程优化，结果表明，应用该方法达到了均衡轧制力和轧制功率的目的，在工程应用中具有很高的实用价值。在轧制规程优化过程中，处理多个优化目标的常用办法是将多个目标加权聚合为单个目标进行优化。为了避免权重赋值的人为因素，对应用多目标进化算法实现轧制规程优化进行了探讨。为了提高算法的执行效率，将量子计算以及混沌与多目标进化算法融合，提出了量子混沌多目标进化算法。标准函数测试表明，量子混沌多目标进化算法的收敛速度比NSGA-II高将近30%，最后将之应用于轧制规程优化，获得了合理的结果，为其在冷连轧机轧制规程优化中的应用提供了理论依据，多目标进化算法是规程优化的未来发展方向。现代化轧机中安装了数以百计的高精度传感器，工艺优化计算机系统的数据库中存储了海量的设备状态数据和轧制过程数据。对数据挖掘工具——支持向量机在轧制力预报中的应用进行了研究和实践。首先整理数据库中的海量数据，建立支持向量机进行轧制力偏差预报的样本库。然后利用样本库中样本对支持向量机进行训练并预报轧制力偏差。最后据此对模型计算所得的轧制力设定值进行修正,冷连轧机的轧制力预报精度提高到5%以内，是轧制规程优化的有效手段。在某厂全连续五机架冷连轧机工艺优化计算机系统开发实践中，基于WinCC组态软件和Microsoft SQL Server2005数据库平台，运用WinCC组件和ANSI-C和VBScript脚本语言，完成了轧制规程计算与优化程序，开发了完善的人机界面监视系统，实现了轧制过程数据和设备状态数据的归档、查询等功能。
[Abstract]:Cold strip production occupies an important position in our national economy. At present, although China's steel production is huge, but the overall low level of production, especially the core technology of cold strip production more in the hands of foreign companies. Therefore, modern production line of cold rolling mill built with independent intellectual property rights, high production of cold rolled steel strip quality, enhance the competitiveness of products in the market is the common goal of domestic researchers. Rolling rules set is one of the core technologies in the cold rolling process, the establishment of a more reasonable rolling schedule is one of the effective ways to improve the quantity of cold rolled steel strip, in this aspect in our country and the international advanced level there is a certain gap, must be more overall, more in-depth research, in order to catch up with the international advanced level. This combination of a plant 1450mm continuous five stand cold rolling mill process optimization computer system development project In practice, the optimization research and practice of rolling regulation based on quantum evolutionary algorithm and support vector machine are carried out in the computer system for the process optimization of cold rolling mill.
The mathematical model of rolling process is based on the rolling schedule optimization, process optimization for different targets can create different objective functions and constraint conditions. The quantum rotation gate angle update strategy is studied, through the improvement of quantum rotation gate angle update query table to enhance the convergence performance of quantum evolutionary algorithm; the minimum energy consumption as objective. The application of improved quantum evolutionary algorithm to optimize the rolling schedule, the total output of the rolling lower than 3%. Calculation procedures applied to a factory five stand cold rolling mill in the production since March 2011, save a lot of energy, to create a huge economic benefits.
In the traditional quantum evolutionary algorithm, application query table is updated on the quantum rotation gate angle must be for specific issues of design inquiry form, universality is poor. In order to overcome this shortcoming, the introduction of particle swarm algorithm and differential evolution algorithm, through the heuristic search in angle space update quantum rotation gate angle, put forward two kinds of hybrid quantum evolutionary algorithm. Through the standard function test shows that the hybrid quantum evolutionary algorithm to enhance the global convergence performance of quantum evolutionary algorithm, improve the versatility, applied to rolling planning optimization, the results show that the application of the method can achieve the balanced rolling force and rolling power, has very high practical the value in engineering application.
In the rolling schedule optimization process, commonly used approach to multi-objective multi-objective weighted aggregation is optimized for a single target. In order to avoid the human factor weight assignment, to achieve the rolling schedule optimization using multi-objective evolutionary algorithms are discussed. In order to improve the efficiency of the algorithm, the quantum computing and chaos and more the algorithm fusion, proposed quantum chaos multi-objective evolutionary algorithm. The standard function test shows that the convergence speed of the quantum chaos multi-objective evolutionary algorithm is higher than the NSGA-II of nearly 30%, finally applied to rolling schedule optimization, and obtained reasonable results, which provides a theoretical basis in the cold rolling mill in the rolling schedule optimization application the multi-objective evolutionary algorithm is the future development direction for optimization.
The high precision sensor installed hundreds of modern rolling mills, process optimization computer system database is stored in the equipment condition data of rolling process and data. The data mining tool -- Application of support vector machine to predict the rolling force of research and practice. At first mass data in the database, create a sample database support vector machine for rolling force deviation prediction. Then use the sample sample of training and prediction of rolling force deviation of support vector machine. Finally, based on the rolling force model calculated set value is corrected, cold rolling mill rolling force prediction accuracy increased to less than 5%, is an effective means of rolling schedule optimization.
In a factory for five stand cold rolling mill process optimization practice of computer system development, WinCC software and Microsoft SQL based on Server2005 database platform, using WinCC components and ANSI-C and VBScript scripting language, completed the rolling schedule calculation and optimization program, developed the man-machine interface monitoring system, realize the rolling process and data archiving the equipment condition data, query and other functions.

【学位授予单位】：燕山大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP18;TG335

【参考文献】