Earthquake action in the design of the bridge belongs to the accidental action.Earthquake is difficult to predict,so the frequency of its occurrence is very small when the structure is in use.Once it occurs,the seismic value is very large and destructive,therefore the risk assessment of the damage caused by the earthquake to the bridge structure becomes the main problem of bridge engineering design.Our country is an earthquake-prone area,and bridges built there have relatively large adjacent spans,of which the quality and stiffness distribution is uneven,due to the limitation of terrain and construction conditions.The existing seismic codes are not applicable to the seismic risk assessment method of the bridge; therefore,it is very important and meaningful to study the seismic risk assessment method.In this paper,a new method for estimating seismic risk probability of bridge seismic risk in earthquake-prone area,which is a combination of neural network simulation and probabilistic finite element method(F-R-M)is established.In the method,firstly,the multiple seismic wave of bridge history analysis and get corresponding seismic response to the structure of the bridge,then RBF neural network is established,with numerical field type,seismic wave energy,seismic wave peak number,the pier moment as input parameters,to train and test the network of RBF.After the success of the training,the Monte Carlo method can be adopted to generate a large number of random numbers with the simulation of neural network to get bridge structures data subjected to random seismic response,and the structure failure probability can be calculated.