基于马氏距离累积量和EMD的结构损伤识别两步法

A two-step method for structural damage identification based onMahalanobis distance accumulation and EMD

结构损伤识别一直是结构健康监测和安全状态评估的热点问题,损伤信息少、信噪比低等因素极大增加了结构损伤识别的难度。基于马氏距离累积量(MDC)和经验模态分解(EMD)提出一种结构损伤识别的“两步法”,首先,利用健康状态监测数据作为参考样本,并利用参考样本的MDC值构造损伤识别向量,其MDC值的均值作为阈值,对待测样本进行初步损伤识别;当监测数据中损伤信息较少、信噪比低,利用直接的监测数据损伤识别困难时,则利用经验模态分解方法将监测数据分解成各阶本征模态函数(IMF),再利用各阶IMF的MDC值构造损伤识别向量,并利用统计学方法对损伤识别向量进行概率密度函数拟合,以概率密度函数95%置信区间的上限值作为阈值对结构进一步损伤识别。通过简支梁数值模拟和工字钢的模型试验验证了该方法的有效性及抗噪性。

Structural damage identification is always a hot topic in structural health monitoring and safe state evaluation. Here, a two-step method for structural damage identification based on Mahalanobis distance accumulation (MDC) and empirical mode decomposition (EMD) was proposed. Firstly, health state monitoring data were taken as the reference sample, and its MDC was used to construct the damage identification vector. This vector’s MDC average value was taken as the threshold to do preliminary damage identification for samples to be tested. When damage identification using directly monitored data was difficult due to less damage information and lower signal-to-noise ratio, the EMD method was used to decompose the monitored data into various intrinsic mode functions (IMFs). Then, MDC values of various IMFs were used to construct damage identification vectors, and the statistical method was used to do probability density function fitting for damage identification vectors. The upper limit of the probability density function within its 95% confidence interval was taken as the threshold to do further structural damage identification. Numerical simulation for a simply supported beam and model tests of I-steel verified the effectiveness and anti-noise of the proposed method.