Vortex-induced Force Model: Radial Basis Function Neural Network for Two-edge Box Girders

Vortex-induced vibration (VIV) is a typical nonlinear fluid–structure interaction phenomenon. Although the semi-empirical model is a theoretical and efficient manner, it is difficult to accurately predict the above nonlinear phenomena due to the incomplete mathematical expressions and uncertain parameters. In this article, the recursive radial basis function neural network (RRBFNN) is proposed to modify the typical vortex induced force (VIF) model. The RRBFNN VIF model was first constructed, and the dynamic equation for calculating VIV response was created, based on the significant nonlinear fluid–structure interaction characteristics of VIV. The aerodynamic forces of the two-edge box girders under various reduced velocities and vibration amplitudes were calculated via computational fluid dynamics (CFD) in order to obtain the mentioned VIF model parameters and verify the efficacy of several RRBFNN learning methods. Ultimately, the amplitude of vortex-induced vibration was predicted using the earlier-mentioned effective training method, and the aerodynamic decoupling method was used to investigate the influence of multiple order aerodynamic forces on the vortex-induced vibration response. As a result, the RRBFNN accurately simulates the aerodynamic force in the time domain and frequency domain. The orthogonal least squares (OLS) and exponential basis function are well-suited to the RRBFNN VIF model. The decisive factor of VIV is the fundamental frequency signal in the aerodynamic force. The second harmonic wave can significantly affect the VIV amplitude, while the influence of the third harmonic wave can be ignored.