Bridge Structural Damage Identification Technology Research

Most of existing civil engineering structures, such as bridges, big dams, high architectures etc., built after the foundation of People's Republic of China, have approached their normal life span. Almost all of these architecture structures are subjected to damage due to external loads, environmental effect, excessive service, natural disaster, initial design defect etc. Structural damage detection and assessment has been becoming a focus of increasing interest in civil engineering field. At present, the study on structural damage detection is still at initial stage and the adopted main approaches are theoretical analysis and numerical simulation, but physical models are scarce. This leads to the yielded theories and methods are not sufficiently applicable for practical engineering application. Aiming at this, this paper focuses on developing effective methods of using wavelet and neural networks to detect the damage of bridge structures due to their extensive applications in civil engineering. The different damage states are set up in a bridge model, a simple-supporting I-section steel beam, through different damage location, different damage quantification, single damage, many damages, etc. In each damage state, the structural dynamic responses are acquired according to the dynamic test. The wavelet packet transform is used to decompose the dynamic response to build damage characteristic vector made up of the wavelet packet component energy, which is sensitive to structural damage. Based on the mathematical analysis of the damage characteristic vector, sensitive damage vector and damage index are defined, which are used to locate damage and quanlify damage effectively. On the other hand, this study makes use of the neural network to detect damage location and damage quantification intellectively of the pre-damage I-section steel beam. The natural network through learning many existing damage specimens, builds a model which can detect the damage location and damage quantification of the un-learned damage cases.Experimental exploration the feasibility and validity of applying wavelet...