【作者】 张华君；

【导师】 权龙； 黄庆学；

【作者基本信息】 太原理工大学 ， 机械电子工程， 2013， 博士

【摘要】 近年来,随着我国钢铁产业的迅速发展,客户对板材的不平度精度的要求越来越高,这就使得各钢铁企业对矫直机的要求也越来越高。国外矫直机生产厂家对其设备的核心技术,比如矫直模型、电液伺服控制等技术进行封锁,而国内生产的矫直设备与国外的同类型先进设备相比,其矫直效率和矫后板材质量仍有相当大的差距,这就使得国内一些大型钢铁生产厂家,由于生产的需要,不得不投入巨资引进国外的先进设备,因此,国产化的新一代强力智能矫直机成为国内矫直机生产家的重要目标。新一代强力智能矫直机的核心技术为多液压缸同步伺服控制技术和智能自学习的矫直模型。本文以国内某钢铁公司轧线上的3300mm四重式十一辊热矫直机为建模设计对象,并参考其它热处理线上的矫直机,建立了新一代强力智能矫直机的多液压缸解耦控制模型及其智能自学习的矫直模型,在建模、仿真与现场实际应用的基础上,做了以下研究。(1)首先对国内外矫直机的已有电液伺服控制传递函数模型进行了详细分析,得到其不足之处以及可以借鉴的地方；分析了现有多缸同步控制解耦方法,得出其解耦方法的优缺点；对已有矫直机模型进行研究,为解决国产化的新一代强力智能矫直机的两个核心技术难题做好了准备。(2)从工程的角度出发,以高频响比例伺服阀的控制电压为输入量,以缸的活塞杆位移为输出量,建立了基于高频响非对称比例伺服阀控制非对称缸的阀控缸传递函数模型。其不同于以往的以伺服阀的阀芯偏置为输入量,以缸的活塞杆位移为输出量的阀控缸系统的传递函数模型,更切合实际。并以此为基础,推导并建立了矫直机的四缸控制模型。(3)设计了“组合给定曲线”,解决了大质量、大惯性系统在启停时的冲击问题。建立了改进的单控制量神经元控制器和多控制量神经元控制器,并证明了多控制量神经元控制器的稳定性,应用粒子群优化算法优化多控制量神经元控制器的权重值。分别对已经建立的矫直机单缸和多缸控制模型进行仿真,结果表明,多缸位移过程同步误差最大值为0.18mm,多缸位移稳态同步误差最大值为0.15mm,多缸速度同步过程误差最大值为0.1mm/s,改进的多控制量神经元控制器对矫直机多缸耦合控制对象解耦的有效性和实用性。(4)通过解析法建立了矫直机的矫直模型,并在此基础上增加了弯辊和边辊对矫直辊压下量的叠加作用计算公式,推导并建立了强化板材的矫直力预估解析式,以实际矫直机的结构参数为依据,针对某种型号的板材,给出了矫直模型运用的算法实例。(5)以模糊控制理论为基础,通过解析模型和矫直工艺专家的知识总结,建立了矫直机的智能自学习矫直模型,并通过矫直过程假设,针对某种板材进行矫后学习,得出了与矫直机操作人员和工艺人员近似相同的学习结果,其入口辊缝和出口辊缝的模糊补偿量均为1.02mm,而工艺人员的补偿量入口辊缝为1.20mm,出口辊缝为0.95mm,而且可以通过比较学习结果与工艺人员经验进一步完善模糊系统参数,从而获得完善的智能自学习的矫直模型,实现实际矫直过程的“无人”化。(6)通过实验室矫直机和工业现场矫直机数据分析可以得到,本文所建立的改进的多控制量神经元控制器与智能自学习的矫直模型完全满足新一代强力智能矫直机的性能要求,实验室矫直机的四缸大行程位移同步过程误差最大值为0.50mm,位移同步稳态误差最大值为0.07mm。工业现场的全液压矫直机的四缸大行程位移同步过程误差最大值为0.5mm,位移同步稳态误差最大值为0.3mm,在矫直板材的过程中各液压缸的最大弹跳量为0.25mm。其性能达到甚至超过国外同类产品的控制精度水平。更多还原

【Abstract】 In recent years, with the rapid development of China’s iron and steel industry, the requirements on the unflatness accuracy of the steel plate for the customers has been getting higher and higher, which makes the quality requirements of the leveler higher and higher for the iron and steel enterprises. The core technology of the leveling equipment, such as the intelligent leveling model and electro-hydraulic servo control etc., has been blocked by the leveler manufacturers abroad, and there is still a considerable gap in the straightening efficiency and straightened plate qulity between domestic leveler and the foreign same type equipment. Which makes the most large domestic steel manufacturers have invested heavily in the introduction of foreign advanced equipment for the production needs. Therefore, the localization of the new generation powerful intelligent leveler has become the top priority of the domestic leveler manufacturers. The core technology of the new generation powerful intelligent leveler is the multi-hydraulic-cylinder synchronous servo control technology and the intelligent self-learning leveling model. In this paper, based on four-heavy-style11-roll thermal leveler of a steel rolling line of domestic some steel producer and with reference to the levelers of other heat treatment line, multi-hydraulic-cylinder decoupling control model and its control algorithms, intelligent self-learning leveling model, of China’s new generation powerful intelligent leveler have been established. Based on modeling, simulation and actual application, the following research has been made.(1)The existing transfer function models of hydraulic servo control system of the domestic leveler and abroad have been carried out a detailed analysis, and their research inadequacies and achievements have been gotten in this article. The existing multi-hydraulic-cylinder synchronous control decoupling methods have been analyzed and their advanteges and disadvantages have been obtained. The leveling models of domestic levelers and abroad levelers have been studied and their inadequacies have been gained. All of above gives a full preparation to solve the two key problems of the localization of the third generation powerful intelligent leveler.(2)According to the engineering practice, the control voltage of the high frequency response proportion servo valve is treated as the input signal and the displacement of the hydraulic cylinder piston is treated as the output signal. And the valve-control-cylinder transfer function model has been founded based on the high frequency response asymmetric proportional servo valve controlling asymmetric cylinder. It is different from a conventional transfer function model in which the spool bias of the servo valve is treated as the input signal. So the transfer function is more realistic in this paper. On the above, the four-cylinder control model has been derived and obtained.(3)A "combined given curve" has been designed and it can reduce the shocks of the system with the large mass and inertia at the start and end of the movement process. An improved single control variable neuron controller and an improved multiple control variables neuron controller have been established. The stability of the multiple control variables neuron controller has been proved and the weight value of the neuron controller is optimized using the particle swarm optimization algorithm. By the simulation of. the established single-cylinder and multi-cylinder servo control object transfer function models, the results show that the multi-cylinder displacement process synchronization error maximum is0.18mm, steady-state synchronization error maximum is0.15mm, speed process synchronization error maximum is0.1mm/s. it has been proved effective and practical for the improved multiple control variables neuron controller to decouple the multi-cylinder coupling control system of the full-hydraulic leveler.(4)The leveling model of the full-hydraulic leveler has been established by the analytic method and on this basis the superimposition of the bending roller and the side roller has been added to the roll reduction of the leveler. And the leveling force estimated formula of the reinforced plate has been deduced and established. Based on the structural parameters of the actual leveler, a algorithm instance for the use of the leveling model has been given for a certain type of plate.(5)Based on the fuzzy control theory, by the knowledge summary of the leveling process expert and the established analytical model, the intelligent self-learning leveling model of the full-hydraulic leveler has been established. By the assumption to the actual leveling process, for a certain type of plate after straightened, the learning result of the intelligent self-learning system comes to the approximate same learning outcome from the operators and craft persons of the full-hydraulic leveler. The fuzzy compensations of the entrance and export roll gap are all1.02mm, and the compensation of the entrance gap is1.20mm and the compensation of the export gap is0.95mm from the craft workers. By comparing the results of the learning system with the craft worker’s experience, a perfect intelligent self-learning leveling model will be obtained and the "no-one" leveling process can be achieved.(6)By the data analysis of the laboratory and field, the requirements of industrial field can fully been met by the new generation powerful intelligent leveler based on the improved multiple control variables neuron controller and intelligent self-learning leveling model in this paper. The displacement synchronization process error maximum is2mm, and the displacement synchronization steady-state error maximum is0.07mm during the four-cylinder big stroke of the laboratoty leveler. The displacement synchronization process error maximum is0.5mm, and the displacement synchronization steady-state error maximum is0.3mm during the four-cylinder big stroke of the industrial site leveler an its maximum bouncing amount is0.25mm during its leveling plate process. Its performance has reached or exceeded the foreign same product.更多还原