基于实测频响函数主成分的在役网架损伤识别方法

AN IDENTIFICATION METHOD OF EXISTING TRUSS STRUCTURAL DAMAGE BASED ON PRINCIPAL COMPONENT ANALYSIS MEASURED FREQUENCY RESPONSE FUNCTIONS

鉴于从实际网架动测中得到的频响函数已受到噪声的污染,会使模态分析出现较大的误差,提出了基于实测频响函数和主成分分析的网架损伤识别方法:用网架实测的频响函数数据作为损伤识别的基本变量,建立损伤识别矩阵,通过主成分分析和变量重构对频响函数进行减消噪处理,利用重构的频响函数前几阶主成分,在低维空间中对损伤信息进行分析、提取,并通过多元控制图,来识别网架的损伤。该方法不需要模态参数,避开了模态参数误差所引起的损伤识别不准问题,不需建立力学模型,对网架边界约束条件和结构型式没有特别的限制。为了验证该方法的可靠性,在试验室完成了足尺网架模型在不同损伤情况下动测试验。结果表明,所提出的损伤识别方法简便可行,结果可靠,尤其对噪声环境下和具有一定非线性网架的损伤识别有良好的适应性。

As frequency response functions (FRFs) measured from a practical truss are contaminated by noise and notable errors will arise in modal analysis, a method of truss damage identification is proposed based on measured FRFs and principal component analysis (PCA). Using measured FRFs as basic variables, the matrix of damage identification is formed, and noise is decreased or removed from the FRFs by using PCA and variable reconstruction. Using the first several order principal components of reconstructed FRFs, the damage information of the matrix mentioned above can be analyzed and extracted in the lower dimension variable space, thus the truss damages can be identified using the multivariate control chart of the first several order principal components. In this process, the modal parameters and the mechanical model of the truss are unnecessary, so identification mistakes arising from the errors of modal fitting can be avoided. Similarly, there is not any special requirement for the truss boundary conditions and its structural form. In order to validate the reliability of the method, a whole size truss is tested in different damage casses. The test results show that the proposed method is feasible and reliable to detect truss damages. Especially, the method has a good applicability for trusses with noise environment and partial non-linear properties.