Research on static finite element model revision based on neural network

A updating method of the structure static model based on the neural network is introduced. It specifies the application of the neural network algorithm in the revision of the static finite element model. It adopts the method of numerical simulation static load experiment to build a BP network model, which a complex non-linear relationship exists between the physical parameters and boundary conditions of a three-span continuous box girder structure and deflections. The finite element calculation which its input data substituted by inverse simulation data is consistent with the assumptions of the `real´ data. It proved the feasibility and practicality of this method.