Modelling Alumina Feeding and Transport in an Industrial Aluminium Reduction Cell Using a Pragmatic Computational Model

Abstract In this work, a pragmatic computational model, which can be employed as a physics-based digital twin, is used to simulate Alcoa’s aluminium reduction cell. The proposed transient model accounts for the evolution of dissolved and particulate alumina in the bath, with options to solve for the tracer distribution and bath temperature. The model also includes a simplified treatment of anode effects and alumina sludging. The bath flow in the model is based on a detailed CFD simulation that is corrected to be mass conserving. The model predictions, using relevant initial conditions and operational settings ( e.g. feeding patterns), are compared with detailed measurements of alumina and tracer during two industrial measurement campaigns. The comparison of the spatial and temporal evolution of tracer predicted by the model matches quite well with the experimental data. The model is able to predict the experimental observations of spatial and temporal variation of alumina by using a sludging coefficient. Comparison between the model predictions and experimental data shows the slow transition (over many hours) between different levels of sludging at various locations in the cell. The model is able to capture the impact of the feeding pattern on the observed alumina distribution. The slow and dynamic process, not treated in the model and hypothesized to be self-feeding phenomena, is also observed to locally (at some locations) increase alumina level in the bath when compared to the simulation predictions. The model has also been used to simulate the evolution of representative bath temperatures in the cell. Despite the simplifications, the model has been shown to be able to reliably model an industrial aluminium reduction cell at a low computational overhead.