Research On Quantitative Damage Identification Method Of Steel Structures With Model Updating

Damage recognition is the key for structural safety assessment,playing an important role in protecting personal and property safety.Traditional damage recognition methods with assumption that the variation of stiffness matrix of damage element is proportion to a single coefficient usually fail to describe the actual degradation within large,complicated structures,although they may behave fabulous in laboratory tests.This paper provides a method for imitate the true behavior in damage structures,and developed a quantitative method for damage identification afterwards.The main results are listed as follow:(1)Based on damage localization with CMCM-MME method,this paper analyzes the influence of the norm value on the results of damage identification in the process of model updating.The results show that when the localization is perfectly conducted,norm value almost had no impact,while the situation changed significantly with the increase of the searching space.For accuracy,searching the best value of p∈[1,2] is recommended to get with unknown damage situations.(2)The reason and the corresponding influence that it is difficult to locate the damage accurately under the assumption of nonproportional damage in three-dimensional space are analyzed.According to the different physical parameters,the stiffness matrix of beam element is proposed to divide it into several sub matrixes to simulate the damage.The stiffness matrix in damage nodes is discussed,which can be represented by a single damage coefficient after simplified calculation.(3)Theory of traditional Genetic Algorithm and Generalized Genetic Algorithm is explained.Applied GGA to the damage identification problem,while the coarse-grained model,mutual exclusion of the same species and self-adaptive mutation probability method is added to help.Combined with CMCM-MME method,a two-stage damage identification method is proposed in this paper.Compared to traditional Genetic Algorithm,GGA performs better in accuracy as well as convergence speed.(4)Three different structure models are used for comparing calculation methods.The results change with different methods,especially for frame structures.The results show that two-stage methods based on GGA can achieve some correction effect when the pre-recognition results have a little problem of wrongly localization,and get relatively reasonable damage identification results.Both scenarios of damage in elements and damage in nodes are analyzed.