Application of the Artificial Neural Network (ANN) in Predicting Anode Properties

Abstract

Carbon anodes are a major part of the cost of primary aluminum production. The focus of the industry is to minimize the consumption of anodes by improving their quality. Therefore, the determination of the impact of quality of raw materials as well as process parameters on baked anode properties is important. The plants have a large data base which, upon appropriate analysis, could help maintain or improve the anode quality. However, it is complex and difficult to analyze these data using conventional methods. The artificial neural network (ANN) is a mathematical tool that can handle such complex data. In this work, Matlab software was used to develop a number of ANN models. Using published data, linear multi-variable analysis and ANN were applied to assess the advantages of custom multilayered feed-forward ANN. Results are presented which show a number of industrial applications.

Comparison of Linear Multivariable, Partial Least Square Regression, and Artificial Neural Network Analyses to Study the Effect of Different Parameters on Anode Properties

“…Normalized value = Value to be Normalized -Minimum Value Maximum Value -Minimum Value [1] Then, the effects of different parameters on the baked anode density were studied using LMA, PLS, and ANN; and the results were compared. Linear multivariable analysis was done using the method described by Bhattacharyay et al [9]. If the input parameters are X1, X2, …, Xn and Y is the output parameter; then, Y can be expressed as:…”

“…The data sets for training were chosen on a random basis. Levenberg-Marquardt training algorithm (LMTA) was used as the back-propagation training algorithm [9]. The advantage of LMTA is that it combines the steepest descent method and the Gauss-Newton algorithm and therefore inherits the speed advantage of the Gauss-Newton algorithm and the stability of the steepest descent method.…”

An artificial neural network model for predicting the CO2 reactivity of carbon anodes used in the primary aluminum production

“…Unlike regression methods, there is no formal rule available for developing ANN models, and it requires a certain level of expertise. Thus, the development of a suitable ANN model is often time consuming [8]. However, the ANN models can make more accurate predictions and give the right trends where the other methods fail.…”

“…Only a few articles [1,8,9] have reported the application of ANN in predicting carbon anode properties. Thus, the scope of ANN in predicting anode properties needs to be explored.…”

“…Bhattacharyay et al [8] compared ANN with the linear multivariable and regression analyses to predict density, electrical resistivity and Young's modulus of baked anodes. It was found that ANN predicts the output for the test data set better than the other two methods of data analysis.…”