Data-driven Aerodynamic Models for Aeroelastic Simulations

Multiple approaches are available for calculating the time-dependent aerodynamic loads of thin, flexible structures subjected to airflow: analytical, semi-empirical, CFD-based, and various reduced-order models. Analytical and semi-empirical models are usually limited for certain structures (e.g., wing profiles, bridge sections, cylinders and prismatic bodies) and small deformations. In the case of large deformations, reduced-order models and CFD simulations are used to calculate aerodynamic loads, but these approaches are computationally costly. We applied a data-based identification method to calculate the aerodynamic loads acting on a flat plate performing heaving and pitching motions. The data-based model was based on high fidelity CFD simulations. The SINDy (Sparse Identification of Nonlinear Dynamics) algorithm was used for model construction. The LASSO (Least Absolute Shrinkage and Selection Operator) and STLSQ (Sequentially Thresholded Least Squares) optimization routines were used to fit the coefficients of the aerodynamic model. The individual models obtained using this procedure have a limited validity range in the reduced frequency of the system. To extend the applicability of the model, linear interpolation was used. The resulting combined model was applied to a two-degree-of-freedom (2-DOF) aeroelastic system. Stability and numerical bifurcation analysis were carried out to investigate the dynamics of the system. The most significant advantage of this method is that the aeroelastic simulations run very quickly after the initial model identification process for a wide range of system parameters and are accurate for large deformation angles.