With the intensive development and utilization of urban underground space, many constructions will inevitably be adjacent to existing subway tunnels, thus forming the complex system of construction disturbance-stratum-existing tunnel. It is necessary to quickly and accurately predict the deformation of this system using cutting-edge technology such as digital twins, to reduce the risk of operation of the urban underground space. In this study, a two-stage surrogate modelling strategy for digital twins is proposed with the help of the predictive capability of the Kriging-based surrogate model and the deformation mechanism of the underground tunnel structure, and it is fully validated using a prototype test. First, a parametric finite element model of this system was established; subsequently, the first-stage surrogate model was trained based on the sample data obtained from the Latin hypercube sampling experiment design for the parametric finite element model. Then, the prediction results of the first-stage surrogate model and the deformation mechanism of the tunnel were utilized to construct the second-stage surrogate model. It is found that the two-stage surrogate modelling strategy improves the prediction accuracy of tunnel deformation from 88.75% to 95.69% as verified by finite element simulation data [1]. Finally, based on the theory of similar modelling, this study designed a prototype test of this system, by choosing reasonable similar material to simulate the effects of construction disturbances on the stratum and existing tunnel structure. By monitoring the deformation process of the stratum and the tunnel, the test reveals how the construction disturbance affects the deformation of the tunnel structure through the deformation of the stratum through the transmission mechanism. As a prototype test of the two-stage surrogate modelling strategy, the prototype test verifies the reliability and practicality of the surrogate model-driven digital twin system through the measured data, which can provide support for the digital twin management of urban underground space.