This paper describes a novel 'hybrid' modelling approach developed in order to predict grain size in a commercial 6xxx series aluminium alloy subjected to a range of different processing and heat treatment conditions. Finite element process simulation models of the various deformation conditions were constructed, and quantitative information on thermomechanical process parameters extracted. This data was used as the basis for analytical models to describe recrystallized grain size in heat treated material. An equation based on the Zener-Hollomon parameter was found to provide some degree of accuracy, however could not account for the large grain sizes observed at high deformation temperature. A modified form of the equation, which included strain rate and temperature in addition to the Zener-Hollomon parameter, exhibited significant improvement in model performance throughout the entire range of grain sizes and an overall predictive accuracy comparable to neurofuzzy data modelling performed on the same results.