ABSTRACT
This chapter presents several recent advances in online decentralized structural identification. First, a dual-rate decentralized approach for online estimation is introduced. At the sensor nodes, the preliminary local estimates are calculated using the raw observations from the corresponding sensor node only. These estimates at each sensor node will be condensed and then transmitted to the base station for data fusion. At the base station, Bayesian fusion approach is proposed to integrate the transmitted local estimates. Two different rates for sampling and transmission are used to reduce the data transfer requirements, so that an efficient online decentralized identification approach is achieved. Then, based on the framework of dual-rate decentralized approach for online estimation, a novel algorithm for online updating of the structural parameters using asynchronous observations from different sensor nodes is introduced. It processes the possibly asynchronous data directly and does not require evaluating the time shifts among the sensor nodes. Finally, a hierarchical outlier detection method is introduced to remove the data anomaly in the decentralized identification. It detects both isolated and segmental outliers according to the definition of outlier probabilities. Through these recent developments, it is expected to achieve efficient and reliable online decentralized identification for structural systems.
