Abstract: Based on enough samples obtained by model experiments in wind tunnel, two BP artificial neural networks (ANNs) were constructed with the MATLAB toolbox of ANN. Then the two ANNs were used to train static coefficients in two different coordinate systems.i.e.body and wind coordinate systems, and training results of static coefficients of bridge section in the two coordinate systems were compared using the Bayesian regularization algorithm. The result shows that a four-layer network is more efficient and has better accuracy. Finally, some problems to the application of ANNs to the identification were pointed out.