Abstract
A digital twin for optimization of the primary alumina feed to the Gas Treatment Center (GTC) has been developed and implemented online at a test plant. The goal of the digital twin is to stabilize the fluoride content of secondary alumina leaving the GTC to generate a stable and predictable raw material for primary aluminium production. The digital twin is based on a first principles process model that predicts fume concentrations, adsorption and H2O and HF transport via gas and alumina flows. The digital twin uses data from the plant and local weather stations and incorporates an estimation routine to adapt the model based on online measurements, including concentration measurements from an HF laser in the main raw gas duct. The digital twin follows the process with high fidelity and is deployed for Model Predictive Control. Results show that the digital twin can stabilize adsorbed fluoride levels in secondary alumina.