Abstract: The total station robot monitoring system is one of the important technical means of subway tunnel safety risk monitoring, but the monitoring points are relatively limited. Based on distributed optical fiber measurement data, the paper proposes a subway tunnel safety risk monitoring technology based on distributed optical fiber. Firstly, the strain data measured by distributed optical fiber monitoring system is mapped to more intuitive deformation and displacement data. Secondly, a strain-displacement transformation model based on BP neural network is constructed by determining 3 input parameters, 2 hidden layers and 1 output parameter. The verification results show that the error between the predicted value and the real value is within 0.2 mm, indicating that the established model can better fit the change law of the data. Finally, it was applied to a certain section of Qingdao metro Line 1, and 401.66～699.45 με is determined as the safe interval of the early warning value of each measurement point, and 202.49～899.98 με is determined as the safe interval of the control value. The calculated results are consistent with the field measured results, and the errors meet the engineering needs.

Key words: distributed optic fiber measurement, BP neural network, strain-displacement conversion model, safety monitoring