Improve the performance of the recursive radial basis function neural network (RRBFNN) model for nonlinear aerodynamic modeling;

An approach to enhance the generalization capability of the RRBFNN is developed based on partical swarm optimization (PSO) or differential evolution (DE) algorithm and model validation;

Proper orthogonal decomposition (POD) is incorporated with system-identification-based RRBFNN to extract more information from the training data;

Multiple activation functions are combined to developing multi-kernel neural networks;

Future work lies in the adoption of advanced maching learning and deep learning strategies to allow modeling systems with varying parameters (multiple operation points like varying Mach number, Reynolds number, or the structure profiles).