Theoretical Study And Application Of Damage Assessment Method Based On Probability Theory

Under the influence of uncertain factors, uncertainty becomes the essential attribute of measurement data and structural analysis model. And the damage detection turns into an indeterminate problem. On the basis of determinate damage assessment method, a new one capable of dealing with uncertainty effectively must be developed in order to identify structural damage in a more appropriate way. At the same time, many approaches developed from inverse strategy fail in the application to complex structure because of difficulty induced by huge solution dimensions. This thesis investigates two key problems that hinder the development of damage assessment, including misidentification and solution difficulty. Stochastic finite element method, super-element theory, probability and evidence theory are introduced into the damage detection procedure established in this thesis. Based on the research of global optimization method, formation of the damage indexes, evidence theory and statistical damage assessment method, a relatively complete statistical methodology of damage location and damage extent identification is developed. The following research results are achieved in this paper:Firstly, on the basis of systemic summing-up of the theory of inverse problem and parameter identification, a regularization-modified parameter identification method is established based on static and dynamic test data. And a new global optimization approach, modified hybrid chaos optimization algorithm, is developed.Secondly, the formation of three damage indexes is carried out based on super-element method and grey correlation theory. The impact of measurement noise and damage extent on the identification of damage location is systemically studied.Thirdly, the damage detection method based on manifold evidences is studied. The evidence theory is employed to extract available information from the results of different damage identification method, then more reliable damagedetection results can be attained by means of evidence combination. A new evidence combination method based on jointed verification of pre- and post- info is established, which takes into account the difference associated with reliability as well as importance of different evidences. The stability and reliability of this approach are demonstrated by experimental results.Fourthly, the stochastic finite element method and probability theory are employed respectively to deal with the influence of different kinds of uncertainty. An automated damage detection method is developed based on statistical damage detection algorithm. The availability of the proposed method is demonstrated through experimental and numerical simulation results. As shown by systemic studies, the method is capable of reducing the probability of misidentification effectively induced by measurement noise.Lastly, a multi-stage damage identification method based on super-element theory is presented for overcoming the solution difficulty of damage identification inherent in complex structures. Then it is demonstrated through experimental and numerical results. The research results show that the approach presented is applicable to large-scale structures. Another advantage that the method has over other approaches is its outstanding robustness, exactness and effectiveness.