Abstract: The influence line is an important parameter of the elastic mechanical state of the bridge structure， which can effectively reflect the resistance and deformation resistance of the structure， and is expected to be used to evaluate and predict the elastic-plastic response during earthquakes. Taking the influence line of a three-span steel plate composite continuous beam bridge as the model correction target， the bridge model correction research is carried out based on BP neural network. With the expectation of Beta distribution as the earthquake damage index， the overtaking probability expression of the bridge model under various performance levels is fitted， and the seismic vulnerability of the continuous beam bridge structure before and after the finite element modification is analyzed and compared. The results show that the relative error between the measured value and the calculated value can be reduced from 38% to less than 10%， and the earthquake damage index of the modified finite element model is lower than that of the initial model. By incorporating the Beta distribution to weight and integrate different performance levels， the structural vulnerability matrix can be transformed into a seismic damage index， thereby accounting for the damage consequences of different failure levels and providing a more comprehensive representation of the seismic performance of the bridge structure.