Using ANN to study VIV of flexible cylinders in uniform and shear flows

Abstract

The ANN model trained on experimental datasets is developed, especially based on the characteristics of flexible cylinder's VIV, the Bayesian regularization back propagation algorithm is employed to train the presented neural network. Nine intuitive physical parameters are selected according to the governing equations of cylinder dynamics. The results show that the neural network trained with intuitive physical quantity can acceptable predictions predict VIV, and the linear regression value is 0.940. In addition, the range of model parameters is limited in the trained neural network, with around 20% error.