Abstract: Variable-fidelity and multidisciplinary optimization approaches were presented and an automatic design system was developed for turbine blade preliminary design to improve design efficiency. The system integrated analysis tools and design teams, as well as data management system, which enabled performing variable-fidelity and multidisciplinary turbine blade optimization design automatically. Low-fidelity model consisted of experiential formulation was used in the initial phase, and high-fidelity model consisted of accurate computational fluid dynamics analysis and structural analysis was used in three-dimensional simulation of optimization phase. According to fidelity of analysis model, design process was divided into three stages to attain the optimum solution gradually, and both design point and off-design point were considered during optimization. As design space was narrowed gradually and researched by design of experiment, computational cost associated with high-fidelity analysis was reduced. Finally, preliminary design for blades of an axial-flow turbine with two stages was performed to validate the system.