Uncertainty Model Updating Of A Curved Cable-stayed Bridge Based On Rbf Neural Network

Finite element model updating is not only an effective guarantee that the finite element model correctly reflects the actual state of bridge,but also a crucial process in establishing an intelligent monitoring system for bridge health.Uncertainty model updating can reflect the disturbance degree of uncertainty in the practical bridge engineering,providing more comprehensive guidance for the actual bridge structure,among which the method based on interval model does not deal with probability knowledge,and needs relatively little prior information,therefore this method is most easily applied to practical bridge engineering.Model updating requires a large number of iterative computations in the process of solving inverse problems,so it is necessary to construct a mathematical substitute model for the finite element model to reduce the amount of calculation.RBF neural network is a common substitute model,which has the feature of strong nonlinear mapping ability,and it has been tested by a large number of deterministic model updating.However,the function form of RBF is complex.Given input intervals,how to get the output intervals accurately is the difficulty when RBF neural network substitution models are applied to interval models.The main contents of this paper include the following aspects:(1)The development process of the finite element model updating is introduced,especially the research status of the uncertainty model updating is emphasized,and the research background and the significance of this paper are clarified,and some popular model updating methods are introduced and compared in recent years,and the advantages and disadvantages of different methods are discussed.(2)The construction process of uncertain model based on RBF neural network is introduced,including experimental design methods and parameter significance analysis.And this paper introduces the method of solving the uncertain model based on genetic algorithm,discusses methods of solving the multivariate function's range with given variable intervals,and puts forward a numerical method based on uniform design idea,finally combines this interval analysis method with the genetic algorithm.(3)According to the measured data in the laboratory,the parameters of the finite element model of the model bridge tower and the model cable-stayed bridge are updated by the methods mentioned above,and then the updating results are tested with ideal results obtained.