Johnston, R.R.M., and Cresswell, P.J.

This paper reports on a CSIRO project involving experiment and modelling which drew on current knowledge with the objective of developing a state of the art, validated, dynamic model of alumina trihydrate precipitation. The model employs the full particle population balance equation implemented in the dynamic package SPEEDUP. It simulates crystallisation using separate rate laws for nucleation, growth and agglomeration, thus allowing the determination of the individual contributions of each of these component processes to the particle size distribution. Model output was compared with batch kinetic data obtained in a bench scale crystalliser with laboratory liquors.

Advance on earlier work lies in the acquisition of a self consistent set of experimental data combined with use of a relatively new dynamic parameter estimation facility available in SPEEDUP. This enabled kinetic parameters to be evaluated from a full set of experimental points, thus giving a better result than hitherto. New experimental data clearly shows that the rate of agglomeration, in this work, is second order with respect to supersaturation, rather than fourth order as previously reported. The evidence suggested that the rate of agglomeration decreased with increasing particle size. An Arrhenius factor was evaluated for nucleation and agglomeration and the data suggest that both component processes have a similar temperature dependence. The effects of mass of seed charge and agitation intensity were also examined and found to be in agreement with earlier reports.

This work illustrates the value of a dynamic model for the analysis and interpretation of complex kinetic data. The flexibility of SPEEDUP is such that recalibration of the model for industrial liquors and reconfiguration of the process circuit to a continuous multi-tank system is a relatively simple task.