Method Research Of Civil Engineering Structure Damage Detection Based On Incompleted Informations

Incomplete measurements for DOFS, measured data errors and model errors are main factors restraining the development and application of structural damage identification. In view of those factors, based on the measured static displacements, mode shapes and natural frequencies including measurement incompleteness and data errors, the dissertation is devoted to study procedures for the structural damage identification using artificial neural network (ANN) and finite element model(FEM).The state of the art of structural damage identification in civil engineering is reviewed and summarized in detail in the present paper. It's found that incomplete measurements for DOFS, measured data errors and model errors are main difficulties for the further development in the fields of structural damage identification. Also, it's realized that structural damage identification based on incomplete measurements and inexact data is the future development. Thus, the research scheme was determined.ANN algorithms have excellent advantages in inverse problems such as pattern recognition, and have been widely used by more and more researchers in the fields of structural damage identification. The key problem for ANN to verify the capability to identify damage depends on the selection of input parameters, that is, whether the index used to localize and quantify the damage is sensitive or not. In this paper, the combined input parameters for ANNs are constituted by static displacements and several low frequencies, which are easily measured and have good precision. Numerical simulations for the localization and.quantification of the damage were performed by using an improved momentum BP neural network. Identification results indicate that the ANNs have good numerical steadiness and robustness even if only a few nodal mode shapes with noise were obtained. Within the allowable error, the identification results were not much susceptible to the errors.The change in the response of static displacement will reflect the change in the structural stiffness matrix, thus it's natural and direct to identify the structural stiffness based on the measurement of static displacement. In addition, considering that the lower frequencies of structures can be tested with high precision and can reflect the global dynamic properties of structures, the combined input parameters to identify the damage localization and extent for ANNs are formed with static displacements and several low frequencies. And then numerical simulations for the localization and quantification of the damage were carried out using the radical basis function (RBF) networks, in which, in the context of limited measurements, much emphasis was laid...