Abstract: The AGARD445.6 solid wing is considered as the object of this study. The fluid-structure interaction method based on CFD and modal superposition is developed, and the aeroelastic responses at various angles of attack and velocities are calculated. Results indicate that with the increase of initial angle of attack, the flutter velocity decreases when the angle of attack is under 7°, and the flutter velocity increases when the angle of attack is between 7° and 10°. Then, at 10° angle of attack, the unsteady aerodynamic reduced order model based on radial basis function neural network is developed to predict the unsteady aerodynamic forces at different velocities and reduced frequencies. The flutter characteristics at 10° angle of attack are predicted using the time domain Runge-Kutta method and the frequency domain VG method. Then the unsteady aerodynamic reduced order model considering the input of initial angle is established to predict the flutter characteristics at different angles of attack. The analysis on the effect of initial angle on the flutter boundary shows that, at small angle, the increase of the generalized force coefficient amplitude ratio with the increase of the angle leads to the decrease of flutter velocity. When the initial angle is greater than 7°, the flow separation region expands, which changes the pressure distribution on the wing surface, leading to the decrease of the generalized force coefficient amplitude ratio of torsional mode, and hence results in the increase of flutter velocity.