Reaserch On The Structural Damage Identification Method Based On The Fuzzy Clustering Algorithm And Time Series Model

Due to the continuous progress of urbanization and the rapid growth of population in large and medium-sized cities,rail transit has been widely promoted for its convenient travel and effective solution to traffic pressure.The environment,operation,and other factors can affect the rail transit during the whole life,which will pose threat to the personal safety of the public,will affect the rail transit.Therefore,it is necessary to carry out structural health monitoring and damage identification for rail transit.The vibration-based damage identification methods have received significant attention during the past decade,which usually can be categorized into two groups:model-based methods and data-based methods.Model-based methods require accurate finite element model for structures,but they are limited by the computation cost of establishing a complex finite element models and bad performance for actual large structures.Data-based methods may avoid such problems and have been widely studied.Among the data-based methods,time series and auto regressive models have become the object of this paper because of its sensitivity to the damage.First,this thseis proposes an unsupervised damage assessment method based on fuzzy clustering analysis and time series,and a probability evaluation method based on the damage threshold.Both of the two methods are based on the coefficients of autoregressive models of the samples.The former method depend on that whether the tested samples from damaged and undamaged satates were clustered into different categories by the fuzzy-Cmeans clustering discriminant analysis to judge whether the structure is damaged.The latter establishs a damage threshold based on 95%guaranteed rate limit of the Mahalanobis distances of the samplses which are not damaged,then the probability of samples beyond the threshold is used to determain the damage.The numerical study shows that these two methods are effective and applicable to detect the damage of the structure,and have good noise resistance.Second,this thesis proposes 4 kinds of indices for damage location,namely the norm?D^M?,dsfc^of,? and ?.The indices norm?D^M?and dsfc^of are definded based on the model coefficients of auto regressive models,while ? and ? are definded with the variance of the residual errors of the auto regressive models.The numerical study shows that all the indices can accurately locate the structural damage and increase with the damage severity.In addition,the indices norm?D^M?,dsfc^of and a have good noise resistance,which is slightly better than ? index.Finally,the experiment of the full-scale model for the subway segment and the numerical experiment of Wuhan JunShan Yangtze Bridge are carried out to investigat the proposed methods.The results show that the proposed methods have good applicability and effectiveness in damage location for actual structure and large scole structure with good noise resistance.Then taking the monitoring project of Wuhan Metro Line 3 as an example,this thesis displays the construction and good application effect of the monitoring system.The proposed methods are applied to the measured accelerations,the result shows that the structure is in a healthy state,in accordance with the actual situation,which is promising to be used for practical application.