Materials which form the surface and subcutaneous layers of an extrudate experience large deformations when they traverse the die land. This, when added to the inhomogeneity caused by the dead metal zone, leads to considerable modifications to the deformation parameters when compared to the remainder of the extrusion. The distribution of structure is therefore greatly inhomogeneous. Reference to both empirical and physical models of the recrystallisation process indicates that nucleation and growth will differ at these locations in those aluminium alloys that are usually solution treated and aged subsequent to the deformation process. Since static recrystallisation has a significant influence on many of the properties of the extrudate, it is therefore essential to provide the methodology to predict these variations. In the work presented, a physical model, for AA2024, based on dislocation density, subgrain size and misorientation is modified and integrated into the commercial finite element method (FEM) code, FORGE, to study the microstructure changes. Axi-symmetrical and shape extrusion are presented as examples. The evolution of the substructure influencing static recrystallisation is studied. The predicted results show an agreement with the experimental measurement. The distribution of equivalent strain, temperature compensated strain rate and temperatures are also presented to aid interpretation. Importantly the properties of hard alloys improve as the temperature of the extrusion is raised. This phenomenon is discussed and theoretically justified. This paper also presents some innovative work where the physically based models, and the Cellular Automata (CA) method, are combined to simulate the static recrystallisation process. The FEM is adopted to provide the initial morphology and state variables for the structure models, such as the equivalent strain, the temperature and the equivalent strain rate. The subgrain size, and dislocation densities are calculated from physically based models and are transferred to CA models to construct the data required to define the initial state for recrystallisation. Simulation results are compared with experimental measurements. It is demonstrated that CA integrated with the physically based models is effective in predicting the structural changes by selecting a suitable neighbourhood and reasonable transition rules.