Aiming at the bridge project in the construction of the development of the status quo of the overdevelopment, maintenance and management level lagging behind, this paper, under the premise of ensuring the safety of the bridge, the bridge surveillance monitoring and risk early warning launched a study to solve the problems of its operation and repair and maintenance. For bridge monitoring and safety monitoring, this paper is based on the vibration acceleration of bridge structure damage identification. On this basis, the damage recognition model constructed by using common neural networks convolutional neural network (CNN), long short-term memory network (LSTM) and deep autoencoder (DAE), and the recognition effect of the three models is compared. This for, for the bridge risk problem, this paper utilizes the Extreme Learning Machine (ELM) and Firefly Algorithm (GSO), constructs the implementation of the GSO-ELM algorithm model for early warning of the bridge safety risk, and the experimental results show that the model proposed in this paper has good effect, which provides support for the future development of the bridge structural safety facilities should be developed in the direction of digitization, automation, and networkization.