Distributed Algorithm For Bridge Health Monitoring Based On Wireless Smart Sensor-Imote2

Recently the importance and necessity of Structural Health Monitoring (SHM) are much realized in civil infrastructure fields. However, the researches and testing on large-scale structures are still in the early stage. Prof. Spencer has shown that the damage on structure is a local problem, which means that we need to mount a dense array of sensors to harvest rich information to orientate those damages. It is a big challenge for traditional wired sensors system. In this case, smart wireless sensors which have realized low-cost and densely installation offer a unique opportunity to large-scale SHM research.According to the way of data process, SHM algorithms can be classified to centralized method or distributed method. Traditional centralized data acquisition and processing schemes that are commonly used in wired sensor systems cannot be employed on large-scale civil structures. The distributed approach proposed by Prof. Spencer is a hierarchical and decentralized scheme which captures local spatial information through coordinated computing between neighboring nodes. This scheme is scalable to a large number of smart sensors for advanced structural health monitoring research.This thesis purposes to make a SHM system with distributed computing strategy for large-scale structures bases on a powerful smart sensor node- Imote2. A steel tubular concrete arch bridge model is designed and built in laboratory according to Guang Hua Bridge in ShanTou; it is the most complicated model in SHM research fields in current. Numerical simulations and experimental testing are implemented respectively on the bridge model the distributed approach results from wireless sensor and wired sensor are compared with them from centralized method for SHM research. All efforts in this thesis are as following:The Imote2 smart sensor system has been developed and employed in testing. The data acquisition, wireless communication and software insert problem are discussed, especially signal time synchronization. Several application codes like stochastic subspace identification are programmed by nesC language in Tinyos to prepare for future work. The efficacy of those codes is demonstrated through several experimental testing.A decentralized modal identification method is proposed, which is fundamental work for distributed computing strategy in SHM. Firstly, Stochastic Subspace Identification which is used to get modal properties of local groups is discussed. Frequency stabilization diagram with singular spectrum entropy derivative is proposed to determine the modal order of testing structure. Vibration testing is simulated on a simple supported beam to compare the modal results of stochastic subspace method from centralized identification and distributed identification respectively to show the efficiency. Then a distributed approach to obtain global mode shape for wireless sensor topology net work is discussed. Particle swarm optimization method is used to get local mode shapes rescale factor, then the global mode shape is combined from rescaled local mode shapes. Using a numerical plate as example shows the distributed method results in different local topology cases. Vibration testing is implemented on the arch bridge model, the vertical bending mode shapes are obtained by distributed method in different local group topology cases, and the results are pleased compared with them from centralized method.The final work in thesis is to detecting link damage on arch bridge model, this is the first time using wireless sensors networks and distributed computing strategy on so much complicated structures. The arch bridge finite element model (FEM) is set up, and then the flexibility curvature method is used to detect link damage which is simulated on FEM in the case of dividing whole FE model into two groups. In the case of having difficult to set up FEM for some real structures, and for comparing with flexibility curvature method results, the difference of power spectral density and curvature difference of power spectral density are used to detecting link damage on FEM as well. Their results are much closed with each other. Finally experimental damage testing is implemented on the arch bridge model. The damages in different case have been made artificially by releasing the bolts of link. Previous distributed methods have been employed to detect those damages through dynamical testing in considering of sensor position arrangement and local groups dividing.