This talk presents a novel hybrid approach, called the Cellular Automata-Phase Field model, which can accurately predict the dendrite formation in a large domain, which combines a 2D/3D CA model with a 1D PF component. In this integrated model, the PF component is employed to accurately calculate the local growth kinetics including the growth velocity and solute partition at the solidification front while the 3D CA component uses the growth kinetics as inputs to update the dendritic morphology variation and composition redistribution throughout the entire domain. The CAPF approach can improve the computational efficiency by 5 orders of magnitude over the PF approach. Using this approach, the prediction of microstructure morphology in a large domain is demonstrated based on the temperature field calculated by the accurate thermal model of additive manufacturing. Case study examples are shown for directed energy deposition of Ti6Al4V alloy and laser welding of SS304.