Study On Structural Damage Identification Method Based On Vibration Characteristic Index And Kriging Model

Structural damage identification is always a hot topic in the field of civil engineering, aerospace and ocean engineering. The theory and experimental technology of damage identification have been unprecedentedly developed for the last half-century. People are focusing on finding a real-time, robust, nondestructive and efficient way for structural damage identification, in which methods of damage identification based on vibration measurement attracts attention and a great number of research in the community of academy and engineering extensively. The basic thought is to detect damage location and predict its level by using the changes in structural dynamic characteristics and responses caused by the damage. Structural damage identification can be seated in the theory of system identification, which doesn’t simply mean to diagnose and repair the structural function and performance, but also makes people reconsider the original structures and help designers improve similar structures further.The damage assessment consists of four sections in the full sense, that is, to estimate whether the damage is present or not, to locate the damage, to do quantitative analysis of the damage and to predict the remaining service life of the structure. In this thesis, the author starts with damage location detection and quantitative analysis. The previous literatures summarized and divided into damage localization, damage quantification, and crack identification. The fundamental research on damage identification method of continuum structural is systematically studied, which can be summarized in the following respects.1. Some kinds of basic index for damage localization in recent research are first summarized. A multi-step decision method for structural damage detection is presented to improve its accuracy and reliability based on the different performance of frequency response curvature obtained at different measurement points on localization results. Numerical example of multi-damaged simply-supported beam is used to illustrate the validity of the presented method. The efficiencies and anti-noise abilities of modal curvature method, flexibility curvature method and the proposed method are analyzed and compared in detail. Finally, experimental examples of a multi-damaged pipeline with local reduction and a single-damaged pipeline with local crack based on experimental model are preformed to verify the correctness and feasibility of the proposed method. 2. A scheme for damage detection in plate which is based upon vibration data only out of damaged structure is studied. According to the tight connection between imaginary part of frequency response function (IFRF) and modal flexibility, a new kind of damage index coupled with gapped smoothing method (GSM) based on uniform load surface formula, namely flexibility of IFRF (FIFRF), is constructed. The effectiveness of model shapes index, uniform load surface index and FIFRF index are analyzed in conditions of different noise levels by the numerical examples. On this basis of FIFRF index, a new damage localization method combined with Thompson outlier analysis and GSM is further presented to filter and eliminate the outlier in detection results caused by measurement noise, so as to achieve success on repressing noise.3. Due to the multi-dimension and multi-extreme of the optimization problem on damage quantification, SCE algorithm is introduced as the basic. Damage quantification can be translated to an optimization problem by establishing objective function of vibration information. To improve the efficiency of damage identification and decrease time of structural analysis, the Kriging surrogate model (KSM) is firstly introduced and combined with SCE algorithm, namely Kriging-SCE method, according to the commonalities between the process of SCE complex evolvement and KSM updating. The original finite element analysis (FEA) can be then replaced by Kriging surrogate model with satisfying precision. An adding point criteria based upon multivariate normal sampling method is introduced to solve the problem on KSM updating during the stagnation of current optimization in SCE complex evolvement. Finally the calculation efficiency of original SCE and Kriging-SCE algorithm is compared by numerical and experimental examples.4. An efficient method of crack identification based on a Kriging surrogate model is presented. The initial samples are used to construct the initial Kriging model which characterizes the relationship between the crack parameters and their corresponding structural dynamic responses, instead of the dynamic constitutive relation for reducing re-meshing process at every iterative step of optimization and the time-consuming FE calculation. To improve the accuracy of the surrogate model around the optimal point and reduce the times of FEA, the Kriging-SCE method is improved, which utilizes the current global optimal point to update the initial Kriging model. The effects of initial sampling size on the identification efficiency and the precision of the identified results are also investigated. This proposed method has certain significance for arbitrary crack identification in a plate.