PREDICTION METHOD FOR VORTEX-INDUCED VIBRATION AMPLITUDE OF STEEL TUBES IN TRANSMISSION TOWERS BASED ON NEURAL NETWORK

In transmission towers, steel tubes with large slenderness ratio are prone to vortex-induced vibration (VIV) under low wind speed. In view of the high cost and time-consuming of traditional wind tunnel test and numerical simulation methods, an efficient VIV prediction method for steel tubes of transmission towers based on neural network is proposed in this paper. In order to obtain the data set, a VIV response analysis method is developed for steel tubes with arbitrary connection joints and geometric sizes. By employing a variety of neural network models (BPNN, PSO-BPNN, RBFNN, GRNN) and performance evaluation method, a prediction method of VIV for steel tubes of transmission towers is established. C-shaped and cross-shaped bolts joints steel tubes are selected as examples, and the VIV amplitudes are predicted. The results show that: Through a comparison with experimental results, the accuracy of the proposed method of VIV for steel tubes is verified, and the relative error of maximum VIV amplitude is 3.84% and 5.87%, respectively for C-shaped and cross-shaped steel tubes, so the proposed method can provide reliable samples for neural network models. After the optimization of hyper parameters through 7-fold for 10 times cross validation, all the 4 types of neural network models show excellent prediction accuracy; the GRNN shows the best generalization ability for both C-shaped and cross-shaped steel tubes with R² values of 0.989 and 0.992, respectively. The GRNN model can well predict the maximum VIV amplitude of C-shaped and cross-shaped steel tubes with different reduced mass damping parameters, and has obvious advantages over CFD methods in terms of computational efficiency.