Research On Overlap Mode And Damage Detection Method For Latticed Shell Structures

Spatial structure is a rapidly developing form in buildings, which has many merits such as strong spanning capacity, beautiful modelling, reasonable stress, big stiffness, light weight, good seismic performance and low cost, and latticed shell structure is one kind of main type. Latticed shell structure has complex geometric shape and dynamic property, which brings many problems on structure parameter identification and damage detection. In recent years, Many earthquakes happened around the world have caused a massive collapse of the building and personnel casualty, which is forcing people push the safety consciousness of major engineering to a higher level. So the study on latticed shell structure monitoring theory and technology is of important theoretic value and practical significance. Combined with the domestic and foreign development situation, detailed research and exploration are carried out on the modal identification of overlap-frequencies structure and damage detection of latticed shell structure, The achievements of this dissertation are as follows:1. The applicability of the curvature mode method in damage detection for latticed shell structure is tested. Curvature mode method in damage detection of beam structure has shown good effect, this index is more sensitive than MAC. But a numerical simulation for latticed shell structure shows that recognition effect of curvature mode is very poor for most damage cases, the conclusion is curvature mode method basically is not appropriate for latticed shell structure.2. Multi-exciting method of identifying overlap modes is presented. Frequency-overlap is the phenomenon that vibration frequencies are overlaped in muti-freedom structure, when the overlap mode shapes of structure is vector subspace constituted by a few of linearly dependent eigenvectors. If the influence of frequency-overlap is not considered when measuring the mode shapes, the overlap modes may be treated as isolated mode which causes mode omission. For modal identification with overlap frequency, the relationship between the acceleration transfer matrices and overlap mode shapes is deduced, that is, a column of the acceleration transfer matrices is the linear combination of overlap vector bases, which still can be regarded as overlap mode. On this ground, multi-exciting method of identifying overlap frequencies in structure is presented. The excitings are imposed on different degrees of freedom and the selection of different excitings is analyzed. A numerical simulation of a three particle structure is done, the results shows that multi- exciting method is feasible. Meanwhile, the orthogonalization formula of overlap mode shapes about quality array is deduced.3. Mode shape characterization using Zernike Moments in latticed shell structures is proposed. In the structural mode recognition,the most widely used method is the modal assurance criterion(MAC), which shows the correlation of different modes but can not represent the details on mode shape features, especially for modal complex latticed shell structures. Mode shape characterization using Zernike Moments(ZM) is studied for latticed shell structures, and the mode shapes are regarded as image functions, and then transformed into Zernike Moments(including amplitudes and phase angles). Through a numerical simulation of a finite element model of a single-layer spherical latticed shell structure, the results shows that the Zernike Moments can reflect the features of mode shapes and recognize overlap modes , that is superior to the traditional MAC.4. A new damage detection method based on Artificial Neural Networks(ANN) and ZM of mode shapes in latticed shell structure is presented. Latticed shell structure is sensitive to parameter, considering the characteristics of ZM, ZM changing forms caused by damage in latticed shell structure is observed, which reveals the space identifiable ability and damage sensitivity of ZM. So, damage can be detected from the damage library established by finite element model using the ANN recognition and matching as well as the fingerprint of ZM amplitudes and phase angles.Through a numerical simulation of a Kiewitt latticed shell structure, the results show that ANN is effective for single member damage detection, which verified the reliability of the damage index of Zernike Moment.