Constellation, le dépôt institutionnel de l'Université du Québec à Chicoutimi

Résumé

Carbon anode is an important component of the electrolytic production of primary aluminum. Anodes are made of dry aggregates (calcined petroleum coke, butts and recycled green and baked anodes) and the binder coal tar pitch. The production of anodes involves the mixing of the raw materials, the compaction of the paste, and the baking of green anodes to produce baked anodes with appropriate properties. An anode property is influenced by the quality of raw materials and the process conditions. It is usually difficult to correlate a given anode property with a particular constituent of a raw material or a particular processing condition because of the lack of a well-established mathematical relationship available for such a correlation. In these cases, the artificial neural network (ANN) can be a useful predictive tool to analyze the effect of a variable on a desired anode property. The analysis provides an insight into the effect of different parameters on anode properties and, in turn, helps improve the quality of anodes. In this article, the ANN approach will be discussed within the context of the analysis of the data on anode production. The results will be presented which show the application of the ANN methods to predict various anode properties.