Numerical And Experimental Study On Modal Analysis And Damage Detection Based On Vibration Response

Large-scale civil engineering structures are prone to be damaged during their service lives,caused by various environmental loads. For the relationship between the vibration characteristicand physical model, we can detect the damage location and extent using the physical model testinformation. Thus, the techniques of nondestructive damage detection based on vibrationresponse are necessary for structures safety evaluation.It is the foundation to get exact vibration parameters for damage detection based onvibration response. So, the basic theory of Eigensystem Realization Algorithm (ERA) andNatural Excitation Technique (NEXT) are introduced firstly. Then, discussion about how todetermine the system order is given. Singularity entropy increment and its derivative areproposed to determine the system order. A simply supported beam is used to demonstrateapplicability. Simulated results have shown that the Natural Excitation Technique (NEXT) andthe Eigensystem Realization Algorithm (ERA) are effective for the structural modal parametersidentification.For the given vibration response, two damage identification methods based on sensitivityanalysis and constrained optimization are presented in this thesis. Firstly, traditional sensitivityanalysis is difficult to use in the practical problem because it is given by the differentialcoefficient. So the 1st and 2nd order sensitivity analysis for frequency and mode shapes in damageidentification are derived combining matrix perturbation theory with finite element technology.For the serious damage extent and the incomplete measurement for DOFs, the perturbationsensitivity iterative method and a two-step method are proposed respectively. All approach areverified to be effective and valuable by numerical simulation, especially for large structures.Secondly, the structural damage detection problem often can be converted into the constrainedoptimization problem. Three different objective functions which comprising of frequency,incomplete model shapes and flexibility are given. A distinguished global search capabilityoptimization algorithm——particle swarm optimization (PSO) is applied. For the incompletemeasurement for DOFs and mass identified parameters of complicated structures, a sensorsoptimal placement program is deduced including the most damage information, and asubstructure distributed method is proposed. A plane 6-bay truss is used to validate the damagedetection techniques.To demonstrate the capability of the proposed approaches, examples of planar truss or beam structures are analyzed numerically. Furthermore, a cantilever beam experimental model is usedto carry out the evaluation of the damage detection algorithm and the ERA. The modeparameters of undamaged and damaged beam are identified from ERA. Then the baseline modelis given from the updated parameters. The final results show that the present algorithms performvery well in spite of the little structural information and measurement inaccuracies.