For active fracture zones,air-mining zones across the conditions of buried pipelines in different service environments,under different conditions of different failure modes under the strain has a difference and buried pipelines in the air-mining zones of the database is more difficult to establish the status quo,directly affecting the buried pipeline on-line monitoring of the early warning accuracy of the problems.In order to establish the correspondence between the health state of buried pipelines and the monitoring strain characteristics,a method for recognizing the state of buried pipelines across working conditions based on two-branch network and domain confrontation is proposed.The method firstly uses a two-branch network to train a recognition model of buried pipelines under active fracture zone conditions; the accuracy of the model is 94.95%,which improves the early warning accuracy of active fracture zones,and then trains a network model of the correspondence relationship between the health state of buried pipelines and the strains in the mining area using domain confrontation based on the linkage of the data characteristics of active fracture zones and the mining area;this method can realize the accurate prediction of the mining area with small samples.Finally,the experiment shows that the warning accuracy of online monitoring is 96.61%,which confirms the effectiveness of this method.