The invention discloses a bridge global damage early-warning method based on a neural network. The method comprises the steps that monitoring data under a healthy condition is collected; natural vibration frequency of a bridge structure is analyzed; a plurality of influence factors influencing the natural vibration frequency are analyzed; a training sample database is established, a BP neural network algorithm is adopted to establish a BP neural network model between all the influence factors and the natural vibration frequency, and the BP neural network model is tested; monitoring data of a bridge under a current condition is collected, and natural vibration frequency is solved and used as a verification sample; the natural vibration frequency of the bridge under the current influence factors is simulated, and a fitting frequency value is obtained; and a healthy sample and the verification sample are established, and hypothesis testing is performed on the healthy sample and the verification sample. The method has the advantages that the natural vibration frequency of the bridge is analyzed according to the healthy monitoring data of the bridge, a complete damage early-warning process is established with the natural vibration frequency being a damage early-warning index, and the applicability and stability of the method are improved in combination with the neural network.