基于高频电阻抗信号与神经网络技术的结构损伤识别研究

为保证结构的安全性和耐久性,宜对结构早期的损伤情况进行健康监测.采用电阻抗方法和人工神经网络技术对钢制薄梁进行了损伤识别的实验研究,但将测得的电阻抗信号都作为神经网络的输入参数则显得不切实际,所以采用主成分分析的降维方法进行实验数据的预处理,降维后包含着最重要主成分的电阻抗信号代替原始数据将作为神经网络的输入参数.研究表明：采用该技术得到的仿真结果与实验观察非常吻合.

Structural Damage Detection Based on High Frequency Electro-mechanical Impedance Signatures and Artificial Neural Networks

To ensure the safety and durability of the structures,a robust state monitoring technique is generally employed to detect incipient damages in the structures.Based on the experiment,an investigation into damage identification for thin steel beams is presented in this paper using electro-mechanical impedance（EMI） signatures and artificial neural networks（ANNs）.The impracticality of using full-size EMI data to feed ANNs as input is noted.The principal component analysis（PCA）-based is adopted for the measured EMI data for its dimension reduction purposes.The compressed EMI data,represented by the principal components,are then used as ANN input variables instead of the raw EMI data.It is shown that the identification results from the proposed method agree fairly well with the experimental observations.