Study on the damage identification of large power transmission tower based on wavelet packet energy and neural network

Based on wavelet packet energy and RBF neural network, the damage identification of principle material of a long span power transmission tower is presented in this paper. The dynamic response signal of displacement-time of the damage and undamaged tower subjected to seismic excitation is obtained. In addition, the wavelet packet energy curvature difference of pre-damage and post-damage states of tower by decomposing and reconstructing signal is given. By selecting the first to the sixth component of wavelet packet energy, the single and multiple damage locations are identified accurately. Moreover, the RBF neural network is applied on damage extent identification of the tower. The results show that the predicted values are in good agreement with the target values, so that RBF neural network can effectively identify the damage extent. From the analysis of damage sensitivity, it can be found that the damage location may still be accurately identified in addition of 10% and 20% random white noise. Consequently, the proposed method of wavelet packet energy has some resistance capacity to noise.