Time Series Model And Sensitivity-based Structural Damage Identification

Time series model is a statistical method to deal with dynamic data,its coefficients and residual errors are sensitive to structural damages and can be directly extracted from structural response.Consequently,time series model has been extensively used for damage identification.The traditional time series model based structural damage identification methods provide limited information about damage location and severity,and structural damage identification is always an ill-posed problem,moreover,the local structural damage is hard to be identified.Based on the support provided by Basic Research Program of China(2016YFC0802002)and National Natural Science Foundation of China(NSFC,51778258),the structural damage identification method based on time series model is studied from the aspects of theoretical analysis,numerical simulation and experimental validation,including the following parts:(1)A structural damage diagnosis method based on time series model and statistical process control is proposed.One-dimensional time series model is used to fit structural vibration response,and the coefficients of time series model are extracted as damage sensitive features.Then,X-bar control chart is built based on the statistical process control technique to identify the presence of structural damage.The effectiveness of the proposed method is verified through its application to the Experimental Phase II IASC-ASCE benchmark structure and the bolt loosening experiment of a tunnel lining structure.(2)To deal with the problem that time series model and statistical analysis based method is hard to provide information about damage location and severity,a structural damage identification method based on time series model and sparse regularization is proposed.The coefficients of time series model are extracted as damage sensitive features from structural vibration response,and the sensitivity of the autoregressive coefficients to structural stiffness reduction factors is derived theoretically.Then,a set of equations for damage identification is established using the changes of autoregressive coefficients before and after damage,and sparse regularization is used to solve the equations.The location and severity of damage are identified using the nonzero entries in the solution,thus the localization and quantification analysis of structural damage is achieved.The effectiveness of the proposed method is verified through the laboratory test of a cantilever beam and the Experimental Phase II IASC-ASCE benchmark structure.(3)Considering that one-dimensional time series model provides limited information about spatial structural properties,a structural damage identification method based on multidimensional time series model is proposed.A multi-dimensional time series model is constructed using the structural vibration response and external forces,and the sensitivity of the multi-dimensional time series model coefficients to structural stiffness reduction factors is derived theoretically.Then,a set of equations for damage identification is established using the changes of autoregressive coefficients before and after damage.Sparse regularization is used to solve the equations,thus the location and severity of damage are identified.The effectiveness of the proposed method is demonstrated by the laboratory test of a six-story lumped-mass shear building structure.(4)To solve the problem that global structural vibration properties are insensitive to local structural damage,a substructural damage identification method based on multidimensional time series model is studied.The large-size global structure is partitioned into several small substructures,and the local part of structure is analyzed independently.A multi-dimensional time series model of the target substructure is constructed using the structural vibration response and external forces.Then,a set of equations for damage identification is established using the changes of autoregressive coefficients before and after damage.Finally,sparse regularization is used to identify the location and severity of damage.The effectiveness of the proposed method is demonstrated by the numerical simulation of a frame and the laboratory test of a six-story lumped-mass shear building structure,and the effect of noise is analyzed.In this paper,the coefficients of time series model are extracted as damage sensitive features,and the sensitivity of the autoregressive coefficients to structural stiffness reduction factors is derived theoretically.Sparse regularization is used to identify the location and severity of damage,and the local structural damages are identified through substructural method.The effectiveness of the proposed methods are verified by numerical simulation and experimental data.