【作者】 李彦荣；

【导师】 彭力；

【作者基本信息】 江南大学 ， 控制理论与控制工程， 2008， 硕士

【摘要】 板带层流冷却控制系统是整个热轧生产线上的一个重要的环节。要完成板带离开精轧机,经过层流冷却系统后达到目标卷取温度的要求,层流冷却系统的控制模型的实施,就必须满足精轧速度和卷取速度的要求,在各生产环节协调的基础上进行。确切地说,水冷区冷却水段的计算公式只能认为是一种理想情况下的静态数学模型。在实际控制中,计算出开阀个数值,并不是立刻就打开相应数目的冷却水段。由于轧制速度的变化,导致板带在输出辊道上的运动是一个变速的过程,为了在实际的复杂工况条件下准确控制板带各点的卷取温度,必须解决动态设定计算、动态跟踪和动态控制的问题。在热轧带钢生产线上,卷取温度的精确控制对带钢质量是至关重要的。本文详细研究了一个实际的热轧带钢卷取温度控制系统。通过研究层流冷却控制系统的工艺,根据国内外应用的一些数学模型的特点,推导并分析了宝钢2050的一阶模型。通过建立了一种简化的动态控制模型,并用一个改进的算法在线调整模型的参数。基于该模型提出了一个包括冷却反馈控制、前馈及自适应联合控制算法的控制器,并对其过程进行了分析,对于卷取温度控制精度主要影响原因进行了进一步的分析。建立精确的热轧带钢卷取温度的数学模型是很困难的,传统的模型需求解复杂的微分方程。此外,数学模型都需要大量的参数来辨识,有些参数不能精确的获得。采用人工神经网络的方式来训练模型参数即模型误差,大大提高了卷取温度的精度。本文采用变梯度BP算法神经网络,即CGBP(conjugate gradient backpropagation)的方法并结合大量的现场数据,对热轧带钢层流冷却数学模型中的模型误差进行预测,将结果应用于计算卷取温度的数学模型中,可以很好的补偿预测带钢的卷取温度,大大提高了卷取温度预测的精度,取得了很好实际应用效果。利用MATLAB仿真程序对模型误差BP神经网络模型进行了离线训练和测试,动态控制模型和CGBP神经网络相结合用于控制卷取温度,分析模型参数和神经网络的预测模型误差的结果是比较理想的。所提出离线仿真结果和在线应用的方法被证明是有效的。通过建立一个更精确的仿真系统,达到指导现场生产的目的。实践结果证明文章中提出的控制方法是有效的,另外仿真系统也具有较大的实际意义。更多还原

【Abstract】 Strip laminar cooling control system is an important process in a hot rolling production line. Left to be completed strip finishing mill, after laminar cooling system to achieve its objectives coiling temperature requirements, the implementation of the control model in laminar cooling system, it must reach the speed and finishing speed coiling at the request of the production chain coordination basis. Properly speaking, the water district cooling water of the formula can only think it is an ideal circumstances static mathematical model. In the actual control, a valve opened calculated value is not immediately open the corresponding number of the cooling water. The rolling speed changes, so that the output roll Strip movement is over the speed of process. Therefore, dynamic set, the dynamic tracking, and dynamic control must be resolved,considering the complexity of the actual working conditions under the precise control panel with the point of coiling temperature.In a hot steel strip production line, the coiling temperature control is critical for strip quality. In this paper, the coiling temperature control of a typical steel strip mill is investigated. Through the research of laminar cooling control system for cooling process, in accordance with domestic and international application of the characteristics of some mathematical model, the first-order model of Baosteel 2050 was derived. A simplified dynamic model is introduced, based on which a cooling control scheme with combined feedforward, feedback and adaptive algorithms is developed.It is quite difficult to model accurately the performance of coiling temperature of hot rolled strip using classical modeling techniques that deal with the solution of complex differential equations. Moreover, the mathematical models require a large number of geometrical parameters to define the system, which may not be readily available. As an alternative, model error can be modeled using an artificial neural network approach with greatly improving the coiling temperature accuracy. By using CGBP (conjugate gradient backpropagation ) neural network , the comprehensive model factor of mathematical model in hot rolled strip was predicted , and the results were applied to calculation of mathematical model of coiling temperature with a good prediction of coiling temperature. The method obviously improves the accuracy of coiling temperature with a good practical application. The CGBP neural networks model is trained and tested by MATLAB. A dynamic model and CGBP neural network are used for control of coiling temperature. Combination of analysis for model parameters and neural network for predicting the error of mathematical model was conducted successfully. Off-line simulation results and on-line application in hot strip mill verify the effectiveness of the proposed method.Simulations with a model validated using actual plant data are conducted, and the results have confirmed the effectiveness of the proposed control method.更多还原