【作者】 贾春玉；

【导师】 刘宏民；

【作者基本信息】 燕山大学 ， 机械设计及理论， 2006， 博士

【摘要】 本文以人工智能理论为基础,选择具有理论和工程实际意义的高精度宽带钢冷轧机板形智能控制为研究课题,进行了深入的理论研究、仿真研究和工业实验研究,取得了新的研究成果。 板形模式识别是板形控制的关键。针对板形模式识别的传统识别方法、模糊识别方法和神经网络识别方法等各自存在的问题,首次建立了以勒让德正交多项式为基模式、以模糊逻辑专家经验知识为支撑、基于遗传-BP算法混合优化、只用3个输入信号、3个输出信号的板形模式识别模糊神经网络模型。该模型不仅网络内部各层节点的物理意义明确,而且自适应能力和抗干扰能力强、识别速度快、精度高,可以满足带钢冷轧机高精度板形控制的要求,为板形模式识别提供了简便实用的新方法。 液压弯辊是板形控制系统最基本的环节,它的动态特性和稳态性能对于整个板形控制系统的性能起着至关重要的作用。针对液压弯辊板形控制系统,建立了电液伺服压力(油压)控制系统的数学模型,制定了一种遗传单神经元自适应模糊控制策略并应用于带钢板形控制中,以提高带钢的成材率,充分发挥液压弯辊力对板形的调整作用,改善轧机系统的动态特性。探索了一种非解析原理的弯辊板形自动控制建模方法,解决了系统建模带来的诸多困难。 板形预报模型是板形控制系统设计的重要基础,无论是板形控制系统中的调节机构控制特性分析,还是在线实时控制,都需要精确的板形预报模型。快速精确的板形预报模型必将提高板形控制系统的控制精度。传统的机理模型通过研究轧制金属内部三维塑性变形和轧辊的弹性变形,建立板形预报模型。由于受到金属本性、轧制条件、轧制设备等多方面因素的制约,同时板形控制系统是一个多变量、非线性、强耦合和纯滞后的控制系统,很难建立其精确的数学模型,因此机理模型难于用在板形在线预报中。为了提升板形预报模型的快速性和准确性,本文以生产实测数据为基础,建立了模糊神经组合式板形在线预报模型。将Elman动态递归神经网络及模糊控制技术引入到了板形预报中,它克服了机理模型中的反复迭代、计算时间长、无法考虑在线动态扰动及多层前馈神经网络存在的易将动态建模变成静态建模问题的缺点,探索了一种非解析原理的板形建模方法,解决了复杂系统建模带来更多还原

【Abstract】 In this paper, author chooses flatness intelligent control for the prevalent high precise cold strip mill as research object, has deeply researched on theory, simulating and industry experiment, and has achieved new fruit.The key of flatness control is flatness pattern recognition. For the problems of the general methods, the method of fuzzy, and method of network for flatness recognition, author has in the first place put forward a fuzzy neural model for flatness recognition using Legendre multinomial as basic pattern, using fuzzy logic expert experience and using genetic-BP algorithm for optimizing, with only three input signals and 3 output signals. The model is simple, fast, precise, self-adapting steady and its nodes have clear physical meanings;so it meets the need of flatness control for cold strip mill, giving easily useful and new method for flatness recognition.Hydraulic bend is basic part of flatness control system. Whether it is dynamic and steady indicates the performance of flatness control system. For hydraulic bend flatness control system, Author has built electric and hydraulic servo pressure (hydraulic pressure) control system algorithm and put forward genetic single nerve cell self-adapting fuzzy control strategy which is used to raise product ratio, make full use of hydraulic force and improve the dynamic performance of mill system. Otherwise, author has explored the method of building bend flatness auto control model using non-parse theory, which overcomes a lot of difficulties in building system model.Flatness predicting model is important to design flatness control system and precise flatness predicting model is needed either in analyzing the control characteristic of the machines adjusted or in controlling on line. Fast and precise flatness predicting mode certainly will raise the control precision of flatness control system. Flatness predicting model is designed by conventional theory model analyzing three-dimensional plastic deformation in rolling metal and elastic deformation of roll. Precise math mode is designed hardly because it is restricted by nature of metal, rolling condition, and rolling equipment etc and flatness control system is many-variable, non-linear, high combined and time-delay;so theory model is hardly used in flatness predicting on line. In order tobuild a more fast and precise model of flatness prediction, author has built on-line fuzzy-neural predicting model based on product data. Elman dynamic recursive network and fuzzy control are introduced in the flatness predicting model and it overcomes a lot of defects, such as iterative operation, time-consuming operation, unable take into account on-line dynamic disturb, easily making dynamic model static model in many-lay ahead feedback network, exploring a new non-parse method of building flatness model and resolving many problems in building complex system model.Because flatness control is many-variable, non-linear, high combined and time-delay, it is difficult to control flatness by use of classical control based on experiential model or modern control. To solve the problem, author has put forward a new intelligent method of flatness control and has built self-adapting fuzzy single nerve cell double-model intelligent control model. According to the combine of fuzzy control and nerve cell control, they can complement each other and make full use of flatness control power. A fuzzy switch pattern is put forward to make the both control signals of the model switch steadily and produce the control signal of the double-model controller, realizing the combine of the advantage of the both control model and improving control performance obviously.Flatness control intelligent model is offered and stimulating software is compiled by combining flatness pattern recognizing model, hydraulic bend control model, flatness predicting model and flatness control model, building flatness control system of cold strip mill and altering the state of controlling flatness by experience.Choosing flatness fuzzy neural network control of high precise wide cold strip mill with theory and engineer practice meaning as research object, author has done a lot of theory, simulating and industry experiment research on flatness pattern recognition, hydraulic bend control, flatness predicting on line, and flatness close loop control. It is important to not only the improvement of flatness high precise control but also the technology of flatness control in theory and engineer application.更多还原