Multi-Models Health Monitoring Method Based On Probability And Structural Identification

Structural health monitoring (SHM) is a powerful tool to understand, evaluate and predict the operating condition of the structure. In this thesis, how to use SHM technologies to study structural performance evaluation is investigated. Two ways are proposed as follows::First, a new flexibility identification method is proposed based on the subspace identification theory to process impact test data. Second, a probability-based multi-model method is proposed to deal with various kinds of uncertainty existing in the process of measurement and structural identification. The main contents and conclusions are as follows:(1)The flexibility identification method is proposed based on the subspace identification theory in the time domain. Unlike the stochastic subspace identification method only identify structure system matrices A and C, the method proposed in this thesis can identify the complete structure system state matrices A,B,C, and D. Therefore, the modal scaling factor can be identified from the proposed method, from which the structure flexibility matrix can be further estimated by utilizing the decoupling characteristics of the complete system state matrix. Once the structural flexibility is identified, it can be used to predict the structure displacement response under any load, it has the potential to substitute the truck static load test for capacity assessment of small bridge.(2) Due to the uncertainty, many FE models are possible to match the monitoring data. To solve the non-unique solution problem, a probability-based multi-models method is proposed. In the proposed method, the Monte Carlo technique and the Markov Chain Monte Carlo technique integrating with Bayesian inference are respectively performed to sample the key structural parameters representing main sources of uncertainty. Then a population of FE models is generated using the samples. Finally, all the models, not the single "best" model, are used for structural identification and performance evaluation.(3) The effectiveness and practicability of the proposed structural flexibility identification method is verified through a three-span girders bridge model and a laboratory 6m simply supported beam.(4) A four-frame benchmark structure is used to verify the efficiency of the probability-based multi-model method, in which the probability-based response prediction and reliability assessment are performed.