As the key parts of an aero-engine, single crystal (SX) superalloy turbine blades have been the focus of much attention. However, casting defects often occur during the manufacturing process of the SX turbine blades. Modeling and simulation technology can help to optimize the manufacturing process of SX blades. Multiscale coupled models were proposed and used to simulate the physical phenomena occurring during the directional solidification (DS) process. Coupled with heat transfer (macroscale) and grain growth (meso-scale), 3D dendritic grain growth was calculated to show the competitive grain growth at micro-scale. SX grain selection behavior was studied by the simulation and experiments. The results show that the geometrical structure and technical parameters had strong inlfuences on the grain selection effectiveness. Based on the coupled models, heat transfer, grain growth and microstructure evolution of a complex holow SX blade were simulated. Both the simulated and experimental results show that the stray grain occurred at the platform of the SX blade when a constant withdrawal rate was used in manufacturing process. In order to avoid the formation of the stray crystal, the multi-scale coupled models and the withdrawal rate optimized technique were applied to the same SX turbine blade. The modeling results indicated that the optimized variable withdrawal rate can achieve SX blade castings with no stray grains, which was also proved by the experiments.