Study On The Related Problems Of Damage Identification Methods For Bridge Structures

Bridge is one of the vital parts of the traffic and transportation system. The healthy status of bridge structures is the premise for traffic and transportation system to operate normally. At present, some bridges appear aging, damage and crack for some reasons, such as design, construction, management, natural disasters and so on. If there is any serious damage in bridges, national economics, social stability, people's lives and properties will be affected. Thereby, it is of great significance in theory and practice to perform health monitoring for bridge structures.Structural damage identification techniques are the important components of health monitoring system of bridge structures. The vibration-based methods for the structural damage identification are the hot researches at home and abroad. Many theories and methods for damage identification have been presented, but the research is inadequate for the structural damage identification methods based on incomplete modal measurements including data noise effect. Considering the state-of-the-art and state-of-the-practice of the structural damage identification methods, the dissertation is devoted to the numerical simulation for the damage identification method of bridge structures with incomplete modal measurements including data noise, on the basis of existing structural damage identification methods. This dissertation includes six chapters as follows:In the first chapter, the background and significance of the dissertation are introduced firstly. Secondly, the development of damage identification and health monitoring for bridge structures is surveyed. And then the state-of-the-art of the structural damage identification methods at home and abroad based on vibration test is summarized in detail and reviewed briefly. Finally, the scope and major research work of this dissertation is expounded.The second chapter is related to the basic theories and several characteristic methods for the structural damage identification.In the third chapter, a damage identification method for bridge structures, based on the residual forces method and improved genetic algorithm (GA), is presented. Firstly, the residual forces method is employed to localize the damage, which can accurately localize the structural damage with any order of the noise-free modal measurements. Noise, however, can not be avoided during the dynamic measurement, which makes the structural damage identification completely impossible using this method. Considering this problem, a new objective function for GA, based on the conventional modal analysis theory, is formulated with the nodal residual forces, and then the damage localization andquantification of the structure with noise is further performed with GA. The improvements on GA include the adoption of real-valued representation, normalized geometric ranking operator, elitist model, arithmetical crossover operator and adaptive mutation operator. Finally, the illustrative examples of a three-bay concrete continuous beam bridge model and a five-bay plane truss bridge model are used to perform the numerical simulation, verifying the efficiency of the presented method. Some research results for GA have been obtained, but several problems, such as the encoding strategy being binary-based, how to choose the initial parameters of GA, how to choose genetic operators and so on, need further to be resolved, which are discussed and investigated in detail and some useful conclusions are made.In the fourth chapter, based on the FEM model and incomplete measured modal data, complete modeshapes are constructed using mode participation factors and a nonlinear damage governing equation is formulated which is applicable to the discrete bar and beam structures, and then a nonlinear least square problem is established, which is solved by the Levenberg-Marquardt least square method. Generally, the nonlinear least square problem is solved by the Gaussian-Newton method, but the problem of structural damage identification is often subjected to the incomplete modal measurements and data noise. To overcome the problems of ill-condition or singularity of gradient matrix caused by Gaussian-Newton method, which results in impossible computation, the Levenberg-Marquardt least square method is adopted to perform simulations on a planar truss bridge model and an arch bridge model, in which the effects of the measured number of degrees of freedom, the number of modes, data noise and the introduction mode of the noise on the structural damage identification are investigated in detail. Results indicate that the method presented shows good performance even in the case of incomplete measurements and limited data noise.In the fifth chapter, a two-step method based on GRNN neural network is studyed, in which some major factors affecting the neural-network-based damage identification method and the corresponding strategy are discussed in detail. The decrease in natural frequency reflects the deterioration in the global stiffness matrix of structures, whereas the change in the mode shape indicates the local change in structures. To make full use of the modal measurements, the combined parameters using a few lower natural frequencies and mode shapes at some nodes, which are easy to measure and have good accuracy, are established and then can be used as the input of neural network after preprocessing. The combined parameters can overcome the shortage of using sole natural frequencies or solemode shapes. The criterion that selects the component of the mode shapes at a few nodes is the elementary modal strain energy coefficient. Because the larger modal strain energy coefficient in the element means the larger magnitude at the corresponding nodes, which is easy to be measured and can be used to train the neural network. Based on the above mentioned, numerical simulation is carried out on a typical truss bridge modal using GRNN neural network, in which the effects of different noise levels with incomplete measured degrees of freedoms and the number of natural frequency on the results of damage identification are studyed.The last chapter summarizes conclusions in the dissertation, pointes out the present problems in the damage identification and health monitoring of bridge structures, and narrates the prospect for the future development of this field.