Abstract: Robots are used to conduct non-destructive defect detection on wind turbine blades and monitor their integrity over time. However, current inspection robots are often bulky, heavy, and struggle to detect defects in the blade’s main beam, thus presenting difficulties in portability and effectiveness. To address these issues, we design a wheel-wing composite robot equipped with a curved surface-adaptive phased array ultrasonic detection device for the detection of defects in the wind turbine blade’s main beam. We determine the pose equation under different section characteristics and identify the robot’s stable range of motion, thus developing a model of its kinematics. A detection device adapted for variable curvature surfaces is designed to ensure tight coupling between the robot’s probe and the blade. Additionally, element differential and least squares ellipse-fitting methods are employed to analyze blades with irregular sections. The simulation results demonstrate that the prototype can stably traverse an area with a vertical angle of ±14.06° at a speed of 0.25 m/s, fully covering the main beam area of the blade during walking operations. Moreover, the robot can scan the main beam area at a speed of 0.1 m/s, enabling the accurate detection of defects.