Abstract: Objective: In the typical scenario of foundation pit excavation adjacent to metro tunnel, a digital twin method for predicting metro tunnel structure deformation based on surrogate model is proposed.Method: A parameterized finite element (FE) model is constructed to simulate the metro tunnel deformation induced by adjacent foundation pit excavation. 13 key parameters are selected as inputs, lateral and vertical displacements at five different positions along the tunnel are taken as outputs. The uniform design is employed to generate FE samples, and a surrogate model is established based on BP (back propagation) neural network. The model is validated using engineering measured results.Result & Conclusion: The predicted lateral and vertical displacements exhibit a precision of over 91.38% in the training and validation sets and over 80.00% in the testing set, with the computation times in the millisecond range. Verification is performed on a foundation pit engineering case in Shanghai Pudong area, where the error between surrogate model predicted and engineering fieldmeasured values is less than 1.5 mm. This indicates that the surrogate model can accurately predict metro tunnel deformation induced by adjacent foundation pit excavation. The above method fulfils preliminary level requirements of accuracy and realtime performance for the application of digital twin in tunnel maintenance.