基于转角模态和小波神经网络的连续梁损伤识别研究

基于小波奇异性检测原理和神经网络非线性映射能力，结合结构基本模态参数，提出了一种结合小波神经网络与结构转角模态的损伤识别方法．首先，建立三跨连续梁的有限元模型获取结构模态参数，并对其进行Mexihat小波变换，通过系数图突变点判断结构损伤位置．然后，将小波系数模特征向量作为BP神经网络的输入，分别研究了该方法在单损伤和多损伤工况下的识别能力．最后将不同工况下神经网络预测值与结构实际损伤程度进行对比，得到单处损伤预测误差平均值为0.22%，多处损伤预测误差平均值分别为0.22%和0.18%，结果表明该方法在结构损伤识别方面的有较高有效性及精确度．

Damage Identification of Continuous Beam Based on Rotation Mode and Wavelet Neural Network

Based on wavelet singularity detection theory and neural network nonlinear mapping ability, combined with modal parameters of structure, a rotation modal based wavelet neural network is established for structural damage identification. Firstly, A finite element model of the three-span continuous beam is established to obtain the structural modal parameters, which are then transformed using the Mexihat wavelet. The locations of structure damage are identified utilizing the discontinuities on the coefficient diagram. Then, the wavelet coefficients modulus eigenvectors are used as the inputs for the BP neural network. The identification ability of the method is studied for both single damage and multi-damage conditions. Finally, the predicted values of the neural network under different conditions are compared with the actual damages of the structure, showing an average error of 0.22% for the single damage condition and 0.22% or 0.18% for the multi-damage condition. The results show that the method is effective and accurate in structural damage identification.