RBF-ARX模型的优化设计及在板形控制中的应用研究

发布时间：2018-08-20 19:44

【摘要】：控制系统的分析和设计中,被控对象或者被控过程的数学模型是极为重要的基础。要控制一个对象或者过程,首先必须要了解其工作机理和特性,其次就要建立其精确的数学模型进行定量分析,最后通过控制达到期望的效果。简单对象或者过程的数学模型容易建立,但是对于一些复杂系统或者过程,其数学模型由于多方面的原因很难建立,所以其控制方案就无法设计。近年来人工智能理论的快速发展,以数据驱动的建模方法引起了广大学者的关注,并将其应用于复杂非线性系统的建模中。本文以RBF-ARX(Radial Basis Function-AutoRegressive eXogenous)模型优化设计及在板形控制系统中的应用研究为课题,在智能控制理论的基础上,设计了遗传算法(Genetic Algorithm,GA)优化的RBF-ARX系统建模方案,并与传统的结构化非线性参数优化方法(SNPOM)优化的RBF-ARX模型做深入的对比研究,实现了板形缺陷模式的识别与控制。首先深入研究了RBF-ARX模型的内部结构和其传统的优化方法SNPOM,针对SNPOM在参数优化过程中运算复杂、占用的存储空间大的缺点,引入了递推最小二乘法进行了改进;同时为了推广RBF-ARX模型在工程领域的应用,进一步提出了GA替代SNPOM的设想,构造了GA-RBF-ARX系统建模和优化方案,大大简化了模型参数的优化过程。其次,针对某900HC可逆冷轧机,建立了GA-RBF-ARX的板形缺陷模式识别模型,仿真验证表明GA-RBF-ARX在板形识别上的效果要比SNPOM优化的效果好。另外,为了验证RBF-ARX模型的建模效果,本文还利用具有联想记忆功能的离散Hopfield网络结构进行了对比研究。最后基于GA-RBF-ARX引入了预测控制策略,建立了包含板形缺陷识别模型、板形缺陷预测模型的完整板形智能控制系统。仿真验证表明GA-RBF-ARX模型在板形缺陷预测上能够跟踪轧机板形的实际输出,同时能够实现板形缺陷的控制精度,满足生产中对板带钢材的要求,是一种有效的建模和控制方法。
[Abstract]:In the analysis and design of control system, the mathematical model of controlled object or controlled process is very important. To control an object or process, first of all, we must understand its working mechanism and characteristics, secondly, we must establish an accurate mathematical model for quantitative analysis, and finally achieve the desired results through control. The mathematical model of simple object or process is easy to build, but for some complex system or process, its mathematical model is difficult to build for many reasons, so its control scheme can not be designed. In recent years, with the rapid development of artificial intelligence theory, the data-driven modeling method has attracted the attention of many scholars, and it has been applied to the modeling of complex nonlinear systems. In this paper, the optimal design of RBF-ARX (Radial Basis Function-AutoRegressive eXogenous model and its application in shape control system are studied. On the basis of intelligent control theory, the modeling scheme of RBF-ARX system based on genetic algorithm (GA) optimization is designed. And compared with the traditional structural nonlinear parameter optimization method, (SNPOM) optimized RBF-ARX model, the pattern recognition and control of shape defect is realized. Firstly, the internal structure of RBF-ARX model and its traditional optimization method, SNPOM, are deeply studied. The recursive least square method is introduced to improve the performance of SNPOM because of its complex operation and large storage space in the process of parameter optimization. At the same time, in order to extend the application of RBF-ARX model in engineering field, the idea of replacing SNPOM with GA is put forward, and the modeling and optimization scheme of GA-RBF-ARX system is constructed, which greatly simplifies the optimization process of model parameters. Secondly, the shape defect pattern recognition model of GA-RBF-ARX is established for a 900HC reversible cold rolling mill. The simulation results show that the effect of GA-RBF-ARX in shape recognition is better than that of SNPOM optimization. In addition, in order to verify the modeling effect of RBF-ARX model, the discrete Hopfield network structure with associative memory function is compared and studied in this paper. Finally, the predictive control strategy is introduced based on GA-RBF-ARX, and a complete shape intelligent control system is established, which includes the shape defect identification model and the shape defect prediction model. The simulation results show that the GA-RBF-ARX model can track the actual output of strip shape in the shape prediction of rolling mill, and can realize the control accuracy of the shape defect and meet the requirements of strip steel in production. It is an effective modeling and control method.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TG334.9

【相似文献】