Research On Bridge Structure Static Finite Element Model Updating Based On Neural Network

In recent years finite element method in structural engineering analysis has been widely used, with the development of modern mechanics, mathematics and computer technology, finite element method has now become the indispensable, structural engineering with wide application effect of numerical analysis tools. During the life cycle of bridge structure the actual state of its static and dynamicresponse values is an issue which is always attended by engineering research field.Estimating responses by finite element program is an efficient method and hasadvantages which are low cost and high speed.But due to every kind of reduction and assumption,the static and dynamic responses which are estimated by the bridge's initial finite element model established with the design drawings often have differences compared with the test results.Therefore,it is an important research direction provided with crucial theoretical and practical value that how to update the initial finite element model and make calculated values approach actual responses of the structure,and the updating technique and results can have application in structure for static and dynamic response analysis,damage identification,health monitoring,and safety assessment and so on.This paper puts forward a kind of make use of universal finite element software ANSYS and neural network combines the model modification method. This paper adopts the method of numerical simulation of based on neural network to the finite element model modification made the feasibility and practicability of validation. Finally, this method is applied to engineering practice, the finite element model of a hyperbolic arch bridge were corrected after correction precision of the model is improved. This paper mainly done the following aspects of work:First:Putting forward the research background and significance of the finite element model correction in this paper,introduction element model updating present study.Second:To introduce the theory and method of finite element model updating. First of all, from the overall to the finite element model modification process, clear idea of model modification; Then attention to detail, respectively from the model parameters selection, the correlation analysis and the establishment of the target function aspects of model modification technology introduction; Finally, A summary of the the finite element model modification technology.Third:To introduce the application of neural network used in model updating. First, an overview of artificial neural network; Derivation of the algorithm of BP network; Combined with engineering practice of using BP neural network for finite element model correction illustrated, in view of the model modification to network input and output parameters is more, Putting forward based on neural network of uniform design parameters optimization method, making sample testing time got reduces and improves the efficiency of the algorithm of neural network.Fourth:MATLAB-based platform, By means of numerical simulating,that whether the parameters can be updated to approach the assumed"actual value"is checked by the model updating method of this paper.The results of simulation indicate that well updating results can be gained, and it also indicates that the model updating method brought by this paper is feasible and practical.Fifth:The proposed model modification method was applied to a curved arch bridge, The load test is done for the bridge before it is operating,and the initial finite element model of the bridge is updated with the method put forward by this paper,after that the updated model is gained at last.The static and dynamic responses calculated by the updated model are even more consistent with the test values than that calculated by the initial model.