Antrobus D., Hiralal I.

The BRAHMA (“BHP-Billiton Refinery Alumina Hydrate Model Analysis”) precipitation simulation model is used extensively at the Worsley Alumina Refinery as a tool to assist in controlling the precipitation process. It combines the method of discretised population balance with the thermodynamic based properties of the supersaturated mother liquor to simultaneously describe the reaction kinetics and the evolution of nucleating, agglomerating and growing aluminium tri-hydrate particles. The dynamic model operates in a closed loop and the precipitation circuit can be easily configured to include features like solids retention, inter-stage cooling, product and seed classification and filtration, green liquor and seed split, etc.

The model provides information on the sensitivity of individual process parameters and the best combination of these process parameters to simultaneously achieve targeted yield, seed/product particle size distribution and occluded soda in product. The model has been configured for the Worsley circuits and validated with field data over a range of operating conditions. This paper relates some of the observations from the calibration process and summarises how the model is being applied in day-to-day operation. Examples of performance improvements and enhanced understanding of the precipitation process emerging from the use of the simulation model are discussed.

Included graphs indicate that alumina trihydrate particles of diameter greater than a critical size do not take part in the agglomeration process. A comparison is made between solid retention precipitators with natural solid density precipitators. The same agglomeration, nucleation and linear growth model parameters have been found to satisfactorily describe the performance of both types.

The model confirms the observed higher overall agglomeration constant in external airlift-agitated precipitators compared to mechanically agitated draft tube tanks. Surface activity for linear growth in mechanical agitated tanks is, however, about 30% higher.