The Optimized Method For Structural Damage Identification Based On Model Updating Technique

The change on physical property causes the change on structural responses.The characteristic has led to many reasearches where the structure responses are utilized for seeking change of local parameters.This is a typical reverse problem in math sense.The natural frequency and mode shape are sturcure responses directly derived from modal test.Besides,many schoolars have propsed several indexes based on modal pairs,such as modal flexibility,curvature mode,stain mode,modal strain energy,etc,which are widely applied in damage identification and model updating field using dynamic data.Expecial studies have been focused on various structures in marine and ocean engineerings,such as the platform for oil exploration,which suffer from extreme service load and would lead to disastrous human and property loss.This dissertation focuses on optimized damage identification methods based on dynamic data and FE model,and propses several improved methods using model updating technique,the key problems,such as optimized sensor placement strategy,is also discussed.The main content of the paper is as follow:(1)Modal flexibility has been widely applied in damage identification and verified to be a modal parameter which is more sensitive to damage,compared with natural frequency and mode shape.This dissertation presents a damage detection method using FE model and flexibility obtained from modal test where the change of modal flexibility before and after damage is defined as the dynamic indicator,and active-set,as well as SQP method is adopted for iterative optimizations.Some of the random factors in real applications,such as measurement noise and model uncertainty,are considered,and Monte Carlo simulations shows statistic knowledge about the effect of these random factors.By mapping the relations of elemental parameters of the original structure into a less-elements substructure,a substructure algorithm is also proposed to lower down the number of unknowns duing iterative calculation,it is verified that this algorithm is capable of improving the accuracy and efficiency for identication results.(2)The sensitivity-based model updating technique has been a hot topic in recent years,and how to determine the design parameters has always been a key problem.Stem from the sensitivity formula of eigenvalues and eigenvactors,this dissertation presents the modal sensitivity formula with respect to any design parameters.It has been verified duing the numerial simulation of a simply-supported beam that the modal flexibility is more sensitive to elemental parameter,compared with natural frequency and the corresponding mode shape.A modal flexibility sensitivity study has been excuted with respect to four distinct element parameters,including elatic modulu,section inertia,mass density and sention area.A single-factor comparison has also been investigated which shows remarkable significance on selecting proper parameters for updating and damage identification purpose based on modal flexibility.(3)Sensor placement is the key of deriving the proper modal data,as well as to model updating and damage detection technique using dynamic data.Begins from the view of observability,this dissertation lists many representive ways of sensor placement and evaluates the goodness of these approaches in three points of view.A minimum root mean square algorithm is proposed based on improving the orthgonality of modes and optimizes the off-diagonal terms in mass-weighted MAC matrix.Two numerical examples have verified the improvement of mode shape orthgonality,and made comparison from view of modal independence and energy.The simulation results has verified that the proposed algorithm is capable of improving the orthgonality of sensor configuration,and some of the advantages are also obtained in view of modal independence and energy.The effect of number of modes and sensors has also been studied,with several important concluions being summarized.(4)CMCM is a model updating approach by directly solving a group of linear equations,which relies on the rank of coefficient matrix to a large extent.Based on the original CMCM method,this dissertation proposes an improved CMCM method.The ICMCM method replaces the business with the difference form in coefficient matix,increases number of effective idependent equations and alters the full-rank condition of the coefficient matix .With ICMCM method,more equations can be obtained by utilizing the same amount mode data and a full-rank preferred coefficient matix is derived,thus the least-square sense solutions can be guaranteed.