桥梁健康监测中的损伤特征提取与异常诊断

针对桥梁健康监测的应用 ,提出了一种基于统计模式识别技术的结构异常诊断方法 .由正常结构实时监测动态响应形成状态判断的数据基 ,通过序列相似分析将未知状态结构的响应信号与正常结构数据基进行环境/运营条件归一化 ;然后根据动态参数模型残差分析提取结构损伤特征 ,进而通过统计分析进行结构异常诊断和损伤定位 .与基于模态频率变化的方法相比 ,该方法具有更高的损伤识别能力 ,并能较好地克服实测数据离散性和环境与运营条件影响带来的困难 .通过三跨连续梁的数值试验验证了该方法的有效性

Damage Feature Extraction and Novelty Detection for Bridge Health Monitoring

A structural novelty diagnosis approach based on statistical pattern recognition techniques is developed for bridge health monitoring.The measured ambient vibration responses of a bridge are employed and the recorded data are divided into many segments to get sets of data samples.For each data sample coming from the structure in unknown condition,a normalization procedure based on hierarchical sequence matching approach is performed to find,in the sample set of the healthy structure,a reference data sample recorded under a similar environmental and operational condition.Then,for each pair of data samples,a damage feature is extracted through autoregressive with exogenous input prediction modeling.Finally,the damage features derived from all the data samples are analyzed statistically and a probability-based measure,damage possibility,is thereby obtained for structural novelty detection and damage localization.The proposed approach is promising for online structural health monitoring and,as illustrated in the case study on a 3-span-girder numerical model,is more sensitive to early-stage damages than modal property based methods.