Cloud numerical computing: Informative support calculation for railway tunnels using data-driven numerical simulation and blocking sub-threaded HTTP protocol

This study introduces a cloud-native framework that alleviates the intricate workflows, limited real-time performance, and heavy software dependency inherent in conventional tunnel-support analyses. A purely data-driven surrogate, trained on 4,933 heterogeneous tunnel sections, replaces time-consuming finite-difference simulations. Latin-hypercube sampling was used to populate the design space and to train a single full-section model that jointly predicts all stability indices. Displacement errors typically fall between 1.3 mm and 2.1 mm, with safety-factor errors around 0.11, while inference is approximately 100 times faster than finite-difference solutions. A lightweight blocking multi-thread HTTP service encapsulates the model, delivering real-time support-design feedback via standard web browsers. The proposed approach lowers the technical threshold of tunnel-support analysis and provides an efficient, real-time numerical computing solution for tunnelling in complex geological conditions.