Efficient modeling of generalized aerodynamic forces across mach regimes using neuro-fuzzy approaches

Abstract

In the presentwork, a nonlinear reduced-order modeling (ROM) approach based on dynamic local linear neuro-fuzzy models is presented for predicting unsteady aerodynamic loads in the time domain. In order to train the input-output relationship between the structural motion and the corresponding flow-induced loads, the local linear model tree (LOLIMOT) algorithm has been implemented. Furthermore, the Mach number is incorporated as an additional input parameter to account for different free-stream conditions with a single model. The approach is applied to the AGARD 445.6 configuration in order to demonstrate the efficiency and fidelity of the proposed method. It is indicated that the ROM-based time domain generalized aerodynamic forces (GAFs) show good agreement with the respective full-order CFD solution (AER-Eu). A further comparison in the frequency domain confirms the validity of the approach. Moreover, the potential of the method for reducing the numerical effort of aeroelastic analyses is highlighted.