铝电解槽极距在线可视化系统的开发与应用

针对铝电解槽极距难以在线测量的特点,将模糊C聚类和多神经网络引入极距软测量模型中,提出一种基于模糊C聚类和多神经网络的铝电解槽实时极距软测量模型。对大量的样本进行模糊C聚类得到优化样本;利用多神经网络对这些样本进行训练得到极距软测量模型;讨论极距软测量模型中辅助变量的确定及其在线获取的方法。在此基础上开发了铝电解槽实时极距在线可视化系统,并在中铝公司广西分公司320 kA系列预焙铝电解槽上进行了安装和调试,系统试运行10个月后,结果表明:造成电解槽不稳定的因素有单振源、多因素耦合和外界扰动;可通过调整单个阳极来消除单振源,提升阳极母线来消除多因素耦合,提高换极和出铝的操作质量来减少外界扰动。

Development and application of real time online visual system of anode-cathode-distance for aluminum reduction cells

The fuzzy C-means and multiple neural networks were introduced into soft sensor model for anode-cathode-distance(ACD) because of the difficulty of online measuring ACD of aluminum reduction cells.Many samples were optimized by fuzzy C-means.The soft sensor model of ACD was obtained by training those samples using multiple neural networks.The determination of auxiliary variables and methods of acquiring them online were discussed.The online visual system of ACD for aluminum reduction cells was developed.The system was installed and debugged on 320 kA prebaked aluminum reduction cells in Guangxi Branch of China Aluminum Corporation.After running the system for ten months,the results indicate that there are three reasons of bringing instabilities of aluminum reduction cells: single oscillating source,coupling of many factors and outside disturbances.Adjusting an anode can eliminate the single oscillating source.Raising busbar of anode can eliminate the coupling of many factors.Improving operation of changing anode and taping metal can weaken the outside disturbances.