Modal Analysis And Damage Diagnosis Of Structure Based On Independent Component Analysis

Civil engineering structures are important part of national infrastructures that are directly related to people's daily life. In recent years, more and more structural health monitoring (SHM) system has been installed in the civil engineering structures such as bridges, long-span space structures, etc. However, as one of the key techniques, the system identification and damage detection has never been ultimately resolved. Therefore, the further research of modal analysis and damage detection method adapting to on-line SHM system of large civil engineering structures has important theoretical significance and practical value. In this dissertation, some new signal analysis techniques such as independent component analysis (ICA), Hilbert-Huang Transform (HHT) have been used for the system identification and damage detection of structures. Some effective means have been proposed, which meeting the need of SHM. The main contents are listed as follow:(1) Based on the introduction of the research state of SHM system, the purpose and significance of this thesis have been explained. The theory significance, the application background, the present development and research results have obtained of structural modal parameter identification and damage detection methods have been commented in detail.(2) The basic principle of ICA, criterion of the independence of random variables have been presented and discussed, as well as the pretreatment methods often used in ICA. More focus has been on the fixed-point algorithm based on negentropy for ICA, which is also called FastICA. ICA is a statistical method for transforming an observed multidimensional random vector into components that are statistically as independent from each other as possible. FastICA algorithm is valid for extracting source component, and works well and stably.(3) The relation between modal coordinates of MDOF system and independent components has been discussed. For free and random vibrations of MDOF systems, modal coordinates have been considered as a specific case of sources that satisfy the relevant assumptions about the source of ICA. To support the previous theoretical findings, numerical experiments of a discrete system of three masses and a free beam have been carried out respectively. The results indicate that ICA is robust to extracting the modal coordinates from dynamic response of MDOF system under free and random vibrations, as well as mode shape vectors. (4) The output-only modal parameter identification method based on ICA has been proposed. Firstly, the modal coordinates and mode shape vectors have been obtained from dynamic response of structure directly by using ICA. Then, the natural frequency and damping ratio of structure have been calculated from modal coordinates by Hilbert transform. For random vibration of structure, the free responses shoud be extracted from independent components by random decrement technique (RDT) before Hilbert transform. Numerical simulation and experimental results indicate that the proposed method is robust to mode shape extraction and suitable for output-only structural modal parameter identification.(5) The structural damage detection method based on curvature mode difference theory has been discussed through theoretical analysis and numerical simulation. The results show that:curvature based damage features are superior to that based displacement; curvature mode is not sensitive to damage located in the mode node; the change rate of curvature mode is more sensitive to damage than curvature mode, whether damage located in mode node or not can be detected.(6) A structural damage detection method based on Hilbert-Huang Transform has been proposed for micro-damage of structure. The ensemble empirical mode decomposition (EEMD) method is used to decompose the modal signal adaptively, the changes of the characteristic frequency are obtained using Hilbert spectrum, and the location of damage are identified effectively.