基于主元分析的网架结构损伤识别方法

提出了利用实测频响函数及主元分析进行网架损伤识别的方法.首先用网架动测得到的频响函数数据建立损伤识别矩阵,用主元分析方法对原始数据变量空间进行降维处理,利用包含原始数据信息最多的前几阶主元,作出多元控制图,进行损伤识别.该方法不需要模态参数,避免了由模态分析中的模态拟合误差所引起的损伤误判.为了验证所提出方法的可靠性,完成了一个足尺网架在不同损伤情况下的动测试验,分析结果表明,所提出的损伤识别方法可行、可靠,尤其对于噪声环境下和具有一定局部非线性的网架有良好的适应性.

A method for truss structural damage detection based on principal component analysis

A method for truss damage detection is proposed based on measured frequency response function(FRF) and principal component analysis(PCA).Damage detection matrix is formed using FRF obtained from truss dynamic test,and its variable space dimensions can be reduced using PCA method,thus truss damages can be detected using multi-variable control chart of first several principal components which contain the most information of all original data.In the process modal parameters are unnecessary,so the mistakes of damage detection,which arise from the errors of modal fitting in the modal analysis,can be avoided.In order to validate reliability of the proposed method,a whole size truss was tested under different damage cases.The experimental results show that the proposed method is straightforward and reliable to detect truss damages.Especially,the method has a good applicability for the truss under the noise environment and non-linear cases.