Research On Damage Identification Method Of Frame Structure Based On Bayes Thought And Statistical Moment Theory

Due to the impact of natural disasters,environmental erosion and other factors,the building structure that has been used for a long time in active service in China is continues deteriorating,resulting in its normal use function being restricted and even causing huge economic losses.Therefore,the whole society is very concerned about the safety of active buildings;and it is significant to evaluate the damage status of the structure to avoid the potential dangers of building structures.However,due to the complexity of the structure and the environment,human and other factors,how to effectively identify structural damage is still a huge challenge.Based on the statistical moment theory of single-degree-of-freedom system,the paper proposes a new method for damage identification in frame structure by combining Bayesian thought and gibbs sampling.By analyzing the sensitivity of different statistical moment damage indexes,the fusion index of fourth-order displacement moment and eigth-order acceleration moment is selected.Combined with the theory of probability density evolution,the application of this method to structure uncertainty can be analyzed.Finally,the feasibility and effectiveness of this method are verified by shaking table test.The main research contents of this article are as follows:Firstly,the theory of statistical moments in single-degree-of-freedom system is derived,and the theoretical formula in multi-degree-of-freedom system is extended.Based on the principles of Bayesian thought and gibbs sampling,a new damage identification method of frame structure is proposed.And then the sensitivity of different order statistical moment is analyzed,three typical damage indicators are chosen.A 12-layer numerical frame model is simulated by considering the signal-to-noise ratio of 40 db and 30 db environmental noise,by comparing the accuracy of the identification results under each damage indicator,the damage indicator with better accuracy and higher noise resistance is selected.After that,the new method and the reference method are compared in terms of identification stability,computational efficiency and anti-noise,and the new method in this thesis is verified to have more advantages.Then the new method is combined with probability density evolution theory,equidistant point selection method and tangent circle point selection method to study the uncertainty damage identification of the same 12-layer numerical frame model.The two random parameters of first damping ratio and amplitude are mainly considered as the single random parameter and the double random parameter to carry out the numerical simulation.The probability density curve of the element elastic modulus is analyzed to verify the application of uncertainty damage identification by using this method.Finally,based on the new method of this thesis,a comparative study on the damage identification effect of a 12-layer standard frame shaking table test is performed.All beam and column elements in 3 typical vibration cases and the typical beam and column elements under the first 20 vibration cases are analyzed for damage identification.Furthermore,the uncertain damage identification results are studied within consideration of probability density evolution theory.It is indicated that the new method can avoid the limitations multiple samplings by using Bayesian analysis,and it can reflect the variation of the damage level of each element with the accumulation of shaking table test cases to a certain extent.It is significant to promote the practical application of the method in the damage identification of frame structure.