The Studies On Structural Damage Detection Using The Method Of Probability Statistics Based On Two-stage Data Fusion

In recent years,with the development of super high-rise buildings、long-span and complex constructions,the study of Structural Health Monitoring has become a popular topic.The defects of design or construction and the external irresistible forces lead to structural damage.The structural damage may add up slowly over time and may cause structural accidents,which result in great loss of lives and property,and impact on the society.Due to environmental noise and incomplete measured mode shapes,modal information based damage detection technology is unable to identify structural damage effectively. This is the bottleneck of such damage detection methods.In this thesis,modal rotations of a structure were obtained by modal translations with the method of static condensation.A probability diagnosis method based on multi-source information fusion was proposed.Firstly,local decisions of structural damage were made with the method of element damage variable and the stiffness reduction factor.And then by using probability and statistics method(including D-S evidence theory and Bayesian statistics theory) to fuse the local decisions, the final identification was made.The simulation results show that this method can reduce the effect of noise and improve the accuracy of damage identification.To avoid the unfavorable influence of the parameter i in the Hankel matrix,the probability diagnosis method based on two-stage multi-source information fusion was put forward.The finite element model of a simple beam was established with ANSYS,and structural acceleration response signals were calculated.The method of probability statistics based on two-stage data fusion was proved to be practicable under different noise conditions by using MATLAB.Dynamic tests were made on a reinforced concrete beam and an I-shape steel beam.The structural vibration acceleration responses of intact and damaged status were measured.The first three modal parameters before and after damage occurred were extracted from the recoded signals.Then the probability diagnosis method based on two-stage multi-source information fusion was used to locate the damage locations.The suggested method was proved to be practical and effective.