Research On Control Methods Of The Concentration Of Alumina Based On Neural Network

Controlling concentration of alumina production is the main method in controlling of the production of Aluminum, so the correct judgment of the concentration of alumina is the basic and key step in the all controlling. But aluminum reduction is an industry process system characteristic of nonlinear, variables-coincident, time transformation and time lag. And it's process accompanied by complicated physical and chemical reaction. Therefore, during the course of aluminum reduction production , the complex production process determined that parameters and variables can't be measured continuously and it have the characteristics of strong-coincidence,indetermination so that it is difficult to establish an accurate mathematical model of recognizing and controlling the concentration of alumina.In recent years, it provided a new method to solve above problems with rapid development of intelligent control theory. In intelligent control method, the neural network has a strong (nonlinear) mapping, self-study and dynamic characteristics ability which can adapt to the indetermination system. Meanwhile, the method allows multiple-input and multiple-output. Then, all of these characteristics give this method a big advantage and enormous potential over another method in solving the nonlinear and indetermination-control problem.According to deeply analyzing on aluminum production and real-time data, article presented a new method for controlling the concentration of alumina. For specific production condition of aluminum reduction, this method aimed at effectively controlling concentration of alumina to keep cell condition and operation work condition of electrolytic cells in good and stable working condition or make massive efforts to improve their production condition.The main contents and achievements of this paper include:(1)Article researches the time series characteristic of Aluminum reduction production. Combining associative rule base, expert knowledge base, control-strategy with ANN, it present a new method of recognizing and controlling the concentration of alumina. New method study and expound the whole controlling model and the connections among ANN model,associative rule base,expert knowledge base particularly, and it makes ANN to be controlling model which can control the controllable parameters(independent variables) of the process of production in real time and in quantification, control concentration of alumina into technological desire ranges, makes status of production to be good all the time, and realizes optimal control in the process of Aluminum electrolytic production.(2) Analyze the existing methods of several typical data flow Trend analysis that including the Sliding Window algorithm (SW) and Extrapolation for On-line Segmentation of Data algorithm (OSD) and their advantages and limitations. Then, by improving the trend analysis method of dynamic data stream, article present a variable sliding window algorithm to achieve a reasonable segmentation for data streams and improve the accuracy of trend analysis. For the real-time data streams of the Aluminum reduction, the method improved the accuracy of trend analysis, thereby enhancing the control accuracy of the entire model(3) Combined with expert knowledge to analyze the affected factors of the concentration of alumina, article use Data warehouse and Data mining technology to do on-line analysis on electrolytic production data. Then, the associative rule of judging concentration of alumina is proposed, and it is a new, exactly, real time method in this field.The method of controlling concentration of alumina ha been effectively utilized in the Gui-Zhou company of ChinaLco in January 2009. The methods can effective and accurately control the concentration of alumina, and achieved the purpose of Energy saving.The feasibility and effectiveness of this approach has been proved by the practical production result.