The Abnormal Diagnosis Of Aluminum Reduction Cell Based On Anode Current

The modern production method of aluminum is electrolysis based on alumina-cryolite.Due to the complex physical and chemical reactions,and the existence of various external fields,the method leads to a complex cell condition.Establishing an effective abnormal diagnosis system can not only improve the quality and increase the yield of aluminum in the production process,but also reduce power consumption.This is important to the production of aluminum.At present,the diagnosis technology of aluminum reduction cell condition has been extensively studied at home and abroad.A abnormal diagnosis method based on analytical model and a abnormal diagnosis method based on intelligence are proposed.Because it is difficult to obtain the data in the process of aluminum electrolysis and there is a certain gap between China and other countries in the Aluminum technics and the Rectifier equipment,some methods can’t be generalized in our country.To solve this problem,a abnormal diagnosis method based on anode current and Neural Network Ensemble.The anode current contains a large amount of information about the cell condition.Through the analysis of the anode current,it can provide the basis for the diagnosis of the aluminum reduction cell.In this paper,we study the anode current by spectrum analysis and extract the characteristic value by calculating the Shannon entropy of each frequency band as the input data of sub-neural network 1.And then we calculate the anode current mean,variance,skewness and kurtosis as the input data of sub-neural network 2.The weighted average of the output of the sub-neural network 1 and the sub-neural network 2 is used as the input data of the decision fusion neural network.Because user’s experience will affect the effect of single neural network and it is difficult to expand for single neural network when there is a new fault or a new eigenvalue,in this paper,we use the neural network ensemble to improve the generalization ability of the diagnosis system.In this paper,we also establish a system for diagnosing the condition of aluminum reduction cell by python.Firstly we collect anode current through the network,and deposited in the SQLite database.And then we preprocess the data,extract the eigenvalues.Finally we diagnose the cell condition by the Neural Network Ensemble.After a series of tests,the system is effective for the diagnosis of the cell condition.