This project draws creative inspiration from the ancient Chinese wooden kite (a wooden bird made of wood, the earliest origin of human kites), breaking through traditional design barriers. By leveraging AI technology to empower the aerodynamic shape design of aircraft in the field of aerospace, it transcends the limitations of human designers' habitual thinking, enhances the efficiency of aircraft design, and opens up more possibilities for innovative breakthroughs in aerospace and other industries. To address the issues of low efficiency, long iteration cycles, and difficulty in breaking through limitations in aerodynamic shape design, this project combines deep learning and aerodynamic optimization. It fully utilizes the learning capabilities of deep learning models to overcome the technical bottleneck of generalized representation of three-dimensional shapes, establishes a dimension-reduction description method for complex three-dimensional shapes, extracts geometric features from cross-sectional images of spacecraft configurations, and achieves effective control of complex aerodynamic layout geometric features and flexible design of any segment. This resolves the mismatch between aerodynamic layout design requirements and current modeling methods.