Forecasting Bath And Metal Height Features In Electrolysis Process

This work presents an initial research on Bath and Metal height features forecasting in electrolysis process based on machine learning multidimensional time series models. It is also examines two clustering architectures in order to find clusters of electrolytic pots with common behavior. The utilized models did not achieve good performance for the Bath height, but in the contrary, the models are able to predict the Metal height within acceptable error margins for the industrial use that is aimed to. An Artificial Neural Network (ANN) and Bidirectional Recurrent Neural Network (BRNN) respectively achieved the best performance for the bath and metal height.