Surrogate modeling for interactive tunnel track design using the cut finite element method

The Cut Finite Element Method (CutFEM) is recently shown to be a versatile approach for tunnel construction modeling and settlement analysis. The CutFEM can significantly simplify the computational workflow by facilitating the discretization and coupling of different components in structural analysis, while maintaining the same accuracy as the standard boundary-fitted method. In this work, three surrogate models based on the Proper Orthogonal Decomposition and Radial Basis Functions (POD-RBF), Artificial Neural Networks (ANN) and Gradient Boosting (GBoost) are proposed and compared for tunnel alignment design. The flexibility of generating multiple designs quickly and conveniently using CutFEM, to provide snapshot data for surrogate model construction, is exploited. As an illustration of the concept, the surrogate model resulting from the linear analysis, involving linear elastic material with compatible parameters, is presented. This approach can be extended to the analysis including a city model, which leads to efficient damage assessment analysis on the existing infrastructure. The resulting surrogate model is used for tunnel track design, where the settlement field can be determined quickly in seconds , which allows real-time applications. The surrogate model is integrated into an interactive platform, named as TunAID, that can be used for real-time design of the tunnel alignment in urban area.