双辊式薄带镁铸轧的参数设定值研究

发布时间：2019-05-23 21:22

【摘要】：双辊式薄带铸轧技术是目前最热门、最有潜力的技术,近几十年这一技术在实验室才得以实现。一些发达国家对双辊铸轧技术的研究处于领先地位,已经率先实现工业化生产。相对于发达国家来说,我国的发展速度较为缓慢,对该技术的研究仍处于实验室生产阶段。双辊式连续铸轧薄带是以液态金属为原料,将其倒入旋转方向相反的两个铸轧辊之间,并以铸轧辊为结晶器,用液态金属直接生产金属薄带的一个完整的生产过程。其工艺特点是将铸造和轧制这两道工序在同一台设备上实现合二为一,与传统热轧工艺相比减少了工序,简化了生产设备,降低了生产成本,节约了能源。因此,这一项技术的研究在工业合金板材生产中十分重要。在双辊式薄带铸轧过程中,有许多参数影响着板材的质量,其中最重要的参数是凝固点的位置,它直接影响了板材的质量。影响凝固点位置的参数有很多,其中最主要的是辊速。在本文中,对铸轧力、辊速以及凝固点位置分别建立数学模型,并推导出凝固点位置与铸轧力、辊速这两个参数之间的关系。根据工艺特点,建立凝固点位置的控制模型,得到铸轧力以及辊速的预设定标定值,以凝固点位置为控制对象,通过实时监测铸轧力的大小,调节辊速的方式将系统凝固点稳定在目标高度。对辊速设定值的预报过程进行了简单的分析,给出预报过程的流程图,为了检验辊速模型的准确性和对系统的适应性,利用实验中心的实际实验数据对其进行验证,比较模型计算出的辊速理论值与实际值之间的差距。对得到的预报值进行PID控制,并通过MATLAB对其进行仿真,验证其可行性。由于在薄带铸轧过程中存在某些参数不可测、某些参数不可控以及环境干扰、系统误差等因素的影响,实际的薄带铸轧生产系统为非线性系统,传统的数学建模适应性较差,因此利用神经网络对确切的数学模型不依赖的优越性,建立一个BP神经网络,取一定量的实验中心的实测数据作为训练样本,对建立的神经网络进行训练,系统输出量与辊速的预报值相对应。为了进一步提高神经网络的预报精度,将之前建立的辊速数学模型计算出的预报值也作为神经网络的一个输入信息,对该神经网络做进一步的改进,同样采用上述的训练样本对其进行训练,训练完成之后,分别再选取实验中心的实测数据作为验证样本对该两种神经网络进行验证,将这两种网络得到的输出信息与实际数据进行对比,得到偏差值。本文以辽宁科技大学镁铸轧工程实验室铸轧系统为实验研究平台,介绍了该实验室的硬件设备以及软件编制情况,分析了部分工艺参数对实验结果的影响。
[Abstract]:Double roll strip casting and rolling technology is the most popular and potential technology at present, which has only been realized in the laboratory in recent decades. Some developed countries are in the leading position in the research of two-roll casting and rolling technology, and have taken the lead in realizing industrial production. Compared with developed countries, the development speed of our country is relatively slow, and the research of this technology is still in the stage of laboratory production. Double roll continuous casting and rolling strip is a complete production process in which liquid metal is used as raw material, which is poured between two casting rolls with opposite rotation direction, and the casting roll is used as mold to directly produce metal strip with liquid metal. The process is characterized by the combination of casting and rolling on the same equipment, which reduces the process, simplifies the production equipment, reduces the production cost and saves energy compared with the traditional hot rolling process. Therefore, the research of this technology is very important in the production of industrial alloy sheet. In the process of double roll strip casting and rolling, there are many parameters that affect the quality of the sheet, among which the most important parameter is the position of the freezing point, which directly affects the quality of the sheet. There are many parameters that affect the position of freezing point, the most important of which is roll speed. In this paper, the mathematical models of casting and rolling force, roll speed and freezing point position are established respectively, and the relationship between solidification point position and casting rolling force and roll speed is derived. According to the characteristics of the process, the control model of the freezing point position is established, and the pre-set calibration values of the casting and rolling force and roll speed are obtained. The solidification point position is taken as the control object, and the casting and rolling force is monitored in real time. The freezing point of the system is stabilized at the target height by adjusting the roll speed. The prediction process of roll speed setting value is simply analyzed, and the flow chart of the prediction process is given. in order to test the accuracy and adaptability of the roll speed model to the system, the actual experimental data of the experimental center are used to verify the prediction process. The difference between the theoretical value of roll speed calculated by the model and the actual value is compared. The predicted value is controlled by PID and simulated by MATLAB to verify its feasibility. Due to the unmeasured parameters, uncontrollable parameters, environmental interference, system error and other factors in the process of strip casting and rolling, the actual thin strip casting and rolling production system is a nonlinear system, and the adaptability of traditional mathematical modeling is poor. Therefore, by using the superiority that the neural network is independent of the exact mathematical model, a BP neural network is established, and a certain amount of measured data from the experimental center is taken as the training sample to train the established neural network. The output of the system corresponds to the predicted value of roll speed. In order to further improve the prediction accuracy of neural network, the predicted value calculated by the previous mathematical model of roll speed is also regarded as an input information of neural network, and the neural network is further improved. The above training samples are also used to train them. After the training is completed, the measured data of the experimental center are selected as verification samples to verify the two kinds of neural networks. The output information obtained by the two networks is compared with the actual data, and the deviation value is obtained. Taking the casting and rolling system of magnesium casting and rolling engineering laboratory of Liaoning University of Science and Technology as the experimental research platform, this paper introduces the hardware equipment and software programming of the laboratory, and analyzes the influence of some process parameters on the experimental results.
【学位授予单位】：辽宁科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TG335.9

【相似文献】