Research On Model Updating And Damage Identification For Frame Structure

The capacity of the structures to resist the extreme load would be decreased in its tens even hundreds of years long service period since the enviorment erosion, material aging, fatigue effect of the alternating load inevitably accumulate the damage and decrease the resistance. So, at present, structure health monitoring become a popular field in the field of civil engineering. And the structure model updating and damage identification are the kernel problem of structure health monitoring.This research work aims at researching the algorithm of model updating and damage identification in the health monitoring field. There are 4 chapters in this thesis:Chapter 1, introduction. The background, objective and significance of the research for this thesis and the domestic and international state of the art are presented.Chapter 2, solid model by ANSYS and plane frame model by MATLAB. A 3-demensional solid concrete frame is simulated using the commercial finite element software ANSYS and the vibration modes are obtained; a simplified plane frame model is established using the software MATLAB and its modes are calculated; the comparison between the modes of the two of them is undertaken.Chapter 3, the model updating using Genetic Algorithm. The plane frame model is updated taking advantage of Genetic Algorithm and the mode results obtained from ANSYS (1st order mode-shape and first 2 order frequencies); The index of modal assurance criterion (MAC) is used to compare the model before and after updating; The participating mass ratio is used to testify that the number of the order of the mode is enough for reflecting the dynamical characters of the solid model.Chapter 4, Damage identification using BP neural network. Take the advantage of two kinds of damage index, ratio of the frequency and the change of the storey displacement in the 1st order mode-shape with neural network. The utilities and the limitations of the two indices are discussed. Since they both work badly under the measurement noise, the reason why noise make damage identification fail is discussed, and point out the better index should be obtained from more accurate measured quantities.Finally, the outcome of this thesis is summarized, and some suggested future work is pointed out.