基于无线智能传感网络的结构损伤识别方法多依赖于集中式的数据采集和处理，具有数据传输量大、基站数据处理负担重、抗干扰和容错性差等局限。本项目利用无线智能传感器独立的信号采集、处理和无线通讯技术，通过理论推导结合数值模型验证，研究考虑无线传感器常见故障的结构损伤在线分布式识别方法。首先，利用非线性滤波方法和贝叶斯数据融合理论，提出多线并行的分布式结构损伤在线双频率识别方法；其次，在分布式计算框架基础上，提出基于离群值概率的多种类异常监测数据在线检验方法；最后，利用基于滑动时间窗口的故障概率模型，提出传感器故障在线诊断方法，并建立基于贝叶斯框架的传感器故障在线重构方法。通过本项目研究可实现解决时变结构传感器故障诊断、重构和损伤识别问题的在线分布式算法，同时能够显著减少数据传输量、减轻基站数据处理负担、提升损伤在线识别效率。

Traditional structural damage detection approaches based on wireless smart sensor networks reply on centralized data acquisition and processing. However, the centralized manner not only requires enormous data transmission and tremendous computational burden at the base station but also has bad anti-interference capability and poor fault-tolerance. The basic feature for smart wireless sensors is the on-board microprocessor, which is used to process the observed data, make decisions, save data locally, and transmit local results. This project takes advantage of smart wireless sensors to establish an online distributed structural damage detection strategy with typical wireless sensor faults. The proposed methodologies will be investigated through theoretical and numerical studies. The nonlinear filter method and Bayesian fusion are employed to develop the online dual-rate distributed structural damage detection approach. Then the hierarchical anomaly detection approach is built using the probability of outlier based on the online distributed identification framework. Finally, the faulty sensors can be identified online using the proposed fault probability, which is established based on moving-time window. Moreover, the faulty sensors can be reconstructed online based on Bayesian framework. The project aims to propose online methodologies for simultaneous faulty sensor identification, reconstruction and structural damage detection. It can resolve the problems of faulty sensors identification, reconstruction and structural damage detection for time-varying structures. On the other hand, this study can significantly reduce the data transmission and alleviate the computational burden at the base station.