论文部分内容阅读
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.
Based on the principle of wavelet singularity detection and the nonlinear mapping ability of neural network, combined with the basic modal parameters of structure, a damage identification method based on wavelet neural network and structural corner modal is proposed.Firstly, the finite element model of three- The structure modal parameters were obtained and subjected to Mexihat wavelet transform to determine the position of structural damage through the mutation point of the coefficient graph.Secondly, the wavelet coefficient model eigenvector was used as the input of BP neural network to study the effect of this method on single and multiple lesions The recognition ability under different working conditions is compared.Finally, the prediction value of neural network under different conditions is compared with the actual damage degree of the structure, the average prediction error of single damage is 0.22%, the average multiple prediction errors are 0.22% and 0.18 %, The results show that the method has high efficiency and accuracy in structural damage identification.