Reduced-Order Model with an Artificial Neural Network for Aerostructural Design Optimization

In this study, an aerostructural analysis using a proper orthogonal decomposition with a neural network is proposed for accurate and efficient aerostructural wing design optimization using the reduced-order model. Because reduced-order-model basis weighting estimation has a limitation in that its robustness cannot be guaranteed by various design variables and wing deformation due to fluid structure interaction, this study employs the neural network, which is capable of perceiving the relationship between the input variables and reduced variables for the proper orthogonal decomposition to complement the defects. To construct the proper orthogonal decomposition with a neural network, the neural network is learned using pairs of design variables and reduced variables from snapshot data obtained from the aerostructural analysis. Because the proposed aerostructural analysis using a proper orthogonal decomposition with a neural network is applied to validation cases and its results are compared to those of the ...