To study the damage identification of bridges using recursive graphs and BP neural networks, taking a certain large cable-stayed bridge as an example, a three-dimensional model was established using ABAQUS finite element software. The acceleration curve of this three-dimensional model was extracted through dynamic analysis and subjected to recursive graph processing and BP neural network analysis. The research results indicate that the recursive graph method can preliminarily identify the location and extent of damage to the main beam. The BP neural network analysis can accurately identify the specific location of damage to the main beam and the degree of damage, with an identification accuracy above 85.0%. This method can provide a reference for damage identification in similar bridge engineering projects.