Research On Methord Of Damage Prognosis Of Cable-Stayed Bridges Based On Validated Model

The cable-stayed bridge is the chief bridge type of modern long-span bridges in our country, but for the lack of experience and knowledge of design, construction and management, together with the severe environment during their service time, damage of cable-stayed bridges stands out as a prominent problem. Based on the finite element model and data acquired by the structural health monitoring (SHM) system, we can simulate the potential damage, and then identify and prognose it, thus giving advice for the bridges’ maintenance and management. Based on the Guanhe Bridge, the main work of this thesis is carried as follows:1. The uncertainty quantization of environment temperature, vehicle load and structure response of Guanhe Bridge is achieved based on data acquired from its SHM system.2. The element model of Guanhe Bridge is modified by Interval Respond Surface Method (IRSM), and the inverse uncertain zone of the model parameter is solved according to monotonicity of the expression of the Interval Respond Surface, completing the model validation.3. Based on the learning algorithm, the relationship between the damage parameter of cables and the structural response of Guanhe Bridge is built in the form of both a explicit RSM formulas and an implicit BP neural network, and the damage is identified and prognosed afterwards.4. Taking the uncertainty of model parameters and vehicle load into consideration, the uncertain fatigue life range of the key fatigue detail of Guanhe Bridge is predicted through method of reliability. The mian results show that:1. The uncertainty of temperature can be expressed by the annual temperature illustrated by a double fourier serial and the diurnal temperature range which conforms to normal distribution; The Gaussian Mixture Model (GMM) can well describe the distribution of vehicle weight while the vehicle-to-vehicle distance obeys a logarithmic normal distribution; A hierarchical probability interval of the structural pesponse can be built through the method of filtering.2. The inversion calculation of model parameter zone can well cover its uncertainty, indicating that the validated model has good function of uncertainty propagation.3. Both the RSM and the BP neural network method can well indentify the random damage degree of the cables, but the BP neural network method shows better accuracy and efficiency; Under the current change trend of cable force, the longest cable in the midspan requires special attention according to the prognosis result using BP neural network method.4. The fatigue life of Guanhe Bridge is between 479 years and 520 years according to the fatigue analysis of its key fatigue detail considering uncertainty of model parameters and imposing a natural vehicle team load; when the linear growth rate of traffic is greater than 10%, attention should be payed to fatigue damage of Guanhe Bridge within its design service life.