A cellular automaton-finite element (CA-FE) model was used to predict the solidification grain structure of per-manent mold cast A7086 alloy. In this model, a Gaussian distribution of nucleation sites was adopted, and the Kurz-Giovanola-Trivedi model was extended to a multicomponent Al–Zn–Mg–Cu alloy system to determine the growth kinetics of the dendrite tip. For describing the Gaussian distribution of nucleation sites on the mold surface, an empirical relationship between the initial cooling rate of the melt and the nucleation density was proposed. Under various casting conditions, the calculated grain structures agreed well with the experimental results. Subsequently, the model was applied to the direct-chill (DC) cast billets, and the simulated grain struc-tures reproduced well the experimental results. This confirmed that the current CA-FE model can be used prac-tically and effectively in DC casting processes.