The Service Life Prediction Of 300 KA Prebaked Anode Aluminum Reduction Cell

The service life of aluminum reduction cell is closely related to the production process,it's an important indicator to measure the quality of electrolytic production process.For the service life of the reduction cell,the existing researches mainly focus on the material quality,few studies analyze the relationship between process parameters and the server life.With the popularization of intelligent management,large numbers of historical data can be stored,how to extract valuable information from data and use the information has a great significance for electrolytic production.Firstly,the characteristics of the history data and the influence of redundant features were analyzed.According to the adverse effect of redundancy feature,feature selection algorithm was used in this paper.Second,for the process parameters,we mainly analyzed the data distribution,trend change,and fluctuation.The cluster analysis was used to analyze the internal structure of the data.Thirdly,multi-scale sample entropy technology was used to analyze the chaos of different electrolytic process.The server life of different reduction cells can be measured accurately based on complexity.The feature submit after processing by mRMR algorithm had a better performance in data cluster.From the view of data distribution,the long life electrolytic cell data is more concentrated.The working voltage data is mainly distributed in the near 3.96 and the fluctuation range of process parameters is smaller.From the point of cell condition,the development trend of the different cells showed that the cell condition is stable in the early stage and complicated in the late stage.The number of cell condition classification of short service life cell was larger than that of long service life of the reduction cell.In the aspect of system complexity,the complexity of different cells is different.The longer life electrolytic cell is less complex than the short.With the increase of the operating time,the complexity of the different cells gradually became stable and different cells can be clearly distinguished.The life prediction based on the complexity characteristics can be used to classify the different life of the electrolytic cell.The results show that the complexity of the electrolysis process is multi-scale;the subset after processing by the mRMR algorithm can improve the accuracy of the data analysis.The extracted feature vector based on the complexity can reflect the characteristics of different cells and can predict the service life.