Structural Parameters And Damage Identification Using Displacement Time Series

Structural parameter and damage identification is one of the most important problems in health monitoring, post-earthquake performance assessment and safety evaluation. It has great significance for maintaining the safety of exsiting structures and normal operation. In the last two decades, many structural parameter identification and model updating methodologies based on eigenvalue and/or mode shape extraction from structural vibration measurement have been proposed. With the development of non-contact sensing technology, it is easy to measure the vibration displacement of the structures. Base on the research of structural health monitoring and parameter identification at home and abroad, the following studies have been carried out in this thesis:(1)A general structural parameter and damage identification strategy using displacement measurement with neural networks is proposed. It is a direct identification methodology only using dynamic response and excitation measurement and no eigenvalue and eigenvectors from dynamic response measurement are needed. The theoretical foundation of the proposed method for structural parameter identification was explained based on the discrete solution of the general equation of motion of structures under dynamic excitation. The strategy can be illustrated by defining an identification index called the root mean square of the prediction difference vector (RMSPDV) and constructing a neural network emulator(NNE) and a parametric evaluation neural network(PENN).(2)Matlab based identification algorithm is developed to implement the proposed method. It has established the filter procedure, the training and creation of training data of neural network emulator and parametric evaluation neural network, and the program for identification.(3)Numerical simulation with a four-story model structure is carried out to validate the feasibility of the proposed approach and the influence of noise to the identification results. Results show that RMSPDV is a good evaluation index for identification with high sensitivity and strong anti-noise ability.(4)The proposed method is verified using test measurement on a two-story model frame structure on a shaking table. The performance of the proposed approach for structural stiffness identification and damage detection using vibration displacement response measurement from laser displacement sensors is verified. Damage is induced by loosening the connection screws in joints of the frame model structure. The parameters identification results are compared with them from traditional identification method based on frequencies extraction. Results show that the inter-story stiffness of the frame structure can be identified with acceptable accuracy. The proposed algorithm is a general and applicable way in practice for parameter identification and damage detection for engineering structures.