利用模态试验参数识别结构损伤的神经网络法

NEURAL NETWORK METHOD IN DAMAGE DETECTION OF STRUCTURES BY USING PARAMETERS FROM MODAL TEST

利用结构位移模态试验和应变模态试验参数和神经网络方法对结构损伤定位和定量辨识问题进行了研究。为获得对结构损伤更加敏感的结构损伤识别指标，在分析现有识别指标的基础上，提出了用于神经网络方法的六种基于结构模态试验参数的损伤识别指标，并对它们进行了实例识别和比较研究。它们均能对结构的损伤进行预报，其中应变类型的损伤识别指标对结构损伤的敏感度比位移类型的损伤识别指标高。

The displacement mode parameters and strain mode parameters are combined withneural networks techniques on the studies of damage detection of structures.In order to obtainmore sensitive signatures for structural damage detection, after considering the analysis ofcurrent signatures, six kinds of signatures for damage detection are studied and the sensitivity todamages is compared. The results show that six kinds of signatures can all be used to predict thestructural damage. It is also clearly seen that damage signatures based on strain data are generallymore sensitive to damage than that on displacement data.