【作者】 杨浩；

【导师】 金晓宏；

【作者基本信息】 武汉科技大学 ， 机械电子工程， 2010， 硕士

【摘要】 高精度轧制是现代轧制技术的一个重要发展方向,高精度轧制技术对冷轧机液压AGC系统的控制精度提出了更高的要求,建立冷轧机液压AGC系统进行的控制系统模型,并对模型的仿真结果进行精度校验具有现实意义。由于仿真技术现在已经成为系统分析、研究、设计必不可少的手段,通过模型仿真可以降低系统的研制成本,尤其是对于不能直接进行试验的系统,系统模型仿真更加凸显出重要性。此外,对于所建立模型的检验也是十分重要的,模型的精确度如何直接关系到模型仿真结果的可信度。本文根据某厂实际生产中的1700mm冷连轧机第一机架实际生产情况,研究分析了相关资料,建立了连轧机组的第一机架控制系统仿真模型,并对仿真结果进行了相应的研究和分析对比。主要内容如下：1)按照实际所采用的监控AGC、前馈AGC和秒流量AGC三种控制策略作为基本控制框架建立的液压AGC辊缝控制的模型主体构架。这三种辊缝控制策略所计算得出的辊缝补偿量相加后输入到液压APC控制系统子模型中,通过位置自动控制来实现消除出口板厚偏差。考虑到现实轧制工况中存在的各种滞后、死区、限幅等环节进行了相应的设置,以便精确地模拟现实轧制过程,建立了AGC控制系统模型的总体控制流程框架。2)针对冷连轧AGC系统存在时间滞后的特点,提出了Smith预测控制算法。该算法用Smith预估器来克服滞后的影响,利用激光测速仪间接测量AGC系统的滞后时间以修正Smith预估器的时滞部分模型。在完成了整个AGC模型系统的建立后,对Smith预估器的控制效果进行了讨论。3)根据具体的液压APC系统,考虑了伺服阀负载流量的非线性和不对称性的基础上建立了三通阀控制不对称缸系统的动力学方程,同时引入负载流量补偿环节进行补偿；按照轧机各部分的弹性变形和轧机负载的质量分布体系,建立了轧机负载系统的模型,同时考虑到冷轧带钢在冷轧时的塑性变形的复杂性,特别引入了轧制力模型,建立了较为全面的液压APC系统的系统子模型。在系统模型仿真部分重点讨论了轧制力和出口板厚这两种输出与实际测量值的对比,通过对比轧制力和板厚时域波形仿真结果和实际测量结果的时域波形,同时计算并对比两种波形其各自的统计特性,得到所建系统模型输出信号的时域波形和统计特性的精确度评价。4)通过对信号的频谱和功率谱分析,可以很快分辨出信号的各个频率成分,本文从频域分析的角度进一步将仿真信号和实测信号进行了研究和对比。5)由于实际工况中需要对大量的非平稳随机信号进行处理,这时需要通过时频分析手段对其进行处理。对于非平稳信号的分析方法分为两类：一类为核函数分解,这类分析方法也称为线性时频描述,如短时傅里叶变换和小波分析等；另一类为能量分布,也称时频能量密度,如Winger-Ville分布、Cohen类等。本文从以下方面就相关的信号进行了分析研究：(a)短时傅立叶变换可以实现一定程度上的时频分析,通过选择合适的窗函数再对信号进行分段的频谱计算就可以得到某一时间区段内频谱随时间的变化情况。小波分析具有可变的时间和频率分辨率,这些性质在理论或应用中都非常重要,作为一种应用十分广泛的信号分析手段可以很好的满足相应的分析需求。本文利用短时傅立叶变换和小波变换各自的优点,分别从时频表述和突变信号的角度给出了模型仿真精度评价。(b)时频能量密度对信号能量的描述是通过时间和频率两个变量来表示,和短时傅里叶变换相比时频能量密度函数具有更好的时频分辨率,同时具有更多的时频分析特性,如真边缘性、有限支撑性、平稳不变性等,是非常有用的非平稳信号分析工具。本文通过对比实测和仿真信号的平滑Winger-Ville分布,从能量分布的角度给出了模型仿真精度的评价。更多还原

【Abstract】 High precision rolling is one of the most important modern rolling technologies which put forward higher requirements to the hydraulic AGC (Auto-Gauge Control) system. System modeling and evaluating simulation results is important and meaningful, and simulation technology has now become an indispensable tool that applied widely in analysis, research and design. Simulation can also make it possible to reduce the cost, especially for the system that can not conduct test directly; hence the importance of simulation is emphasized with these advantages. Furthermore, model testing is equally important because the accuracy of the simulate result cast direct influence to the credibility of the model.This article was accomplished based on the research of real product situation of 1700mm strip rolling mill and analysis of corresponding references, a system control model was established accordingly and the results of simulation were analyzed from different perspectives, the main contents are as follows:1) According to the actual control strategies employed in the rolling product, Monitor AGC, Feed Forward AGC and Mass Flow AGC control blocks were created and the main framework of control system model was established. Roll gap compensation was calculated by these control strategies, so APC system can functioned to reduce the thickness deviation. Lag, dead zone and limiting were added in the control model to make it more accurate.2) Smith predictive control algorithm was applied to reduce the negative influence created by the lag of the system. Time-delay estimate Smith model was corrected by using laser speedometer to measure the lag time indirectly. Control effect of the predictor was discussed after the AGC model was completed.3) According to the concrete hydraulic APC (Auto-Position Control) system, taking the nonlinearity and asymmetry of servo-valve load flow into account, the model of asymmetry cylinder controlled by the three-way servo-valve was establishes. At the same time the load volume of flow compensated loop was added which functioned as compensation in the situation. Taking the elasticity deformations of various parts and the mass distribution system of rolling mill loading situation into account, dynamic system of rolling mill was built. Complexity of plastic deformation for the cold rolling steel strip was considered and the rolling force model has been induced. Based on the mentioned sub models, the thesis presents the mathematical model of hydraulic APC system, which is comparatively complete and accords with the physical device system. The accuracy of simulation results were analyzed by comparing the time domain waveform and the statistical characteristics, and then the evaluation in this aspect for the accuracy of model was obtained.4) Various frequency components of the signal can be identified based on the signal frequency and power spectrum analysis. The simulation signal and the measured signals were studied and compared from the angle of frequency domain analysis.5) Because large amounts of non-stationary random signal were generated during rolling, time-frequency technology was applied to analyses these signals. Non-stationary signal analysis methods are divided into two types:one is kernel function decomposition, such as Short-time Fourier Transform and Wavelet Transform; the other is energy distribution, such as Winger-Ville distribution and Cohen class. The corresponding research are as bellows:(a) Short-time Fourier transform can achieve a certain degree of time-frequency analysis, spectrum within a certain time period can be obtained by calculating spectrums of segmented signal if appropriate window function were selected. Wavelet analysis has variable time and frequency resolution which was very important both in research and application. By using the advantages of Short-time Fourier Transform and Wavelet Transform, accuracy of model was given.(b) Time-frequency energy density of the signal energy is described by time and frequency and time-frequency resolution is better than Short-time Fourier Transform. Mean while, the time-frequency energy density analysis has more features, such as the true marginal, limited supportive, stable invariance, so it’s an efficient non-stationary signal analysis tools. Accuracy of model was given after comparing the measured signal and simulation signal smooth Winger-Ville distribution.更多还原