Flutter instability boundary determination of composite wings using adaptive support vector machines and optimization

Recently, machine learning tools have been vastly used in different engineering problems. This paper determines the composite wing aeroelastic instability boundary by an adaptive support vector machine in which the informative training samples are sequentially selected based on an active learning approach. The wing model is a cantilever beam with two degrees of freedom and the addition of the thrust of the engine as a follower force and a concentrated mass to simulate the engine's inertial force. For structural modeling of the composite wing, the layer theory has been adopted, and in the aerodynamic model, the flow has been assumed to be unsteady, subsonic, and incompressible. The adaptive support vector machine capability in determining the flutter instability boundary has been assessed. The results show that the adaptive support vector machine outperforms conventional support vector machines in accuracy and computational cost. Moreover, a novel optimization algorithm is proposed to maximize the flutter speed of the multi-layer composite wings using the trained support vector machines considering fiber angles as design variables. The results indicate that utilizing the support vector machine expedites the optimization process enormously.