This communication presents some results from a case history in Barcelona involving a large cut-and-cover railway tunnel requiring a deep excavation (depth > 20 m). Prediction of excavation-induced movements was necessary to assess third-party risk. A numerical model was set-up for the prediction and extrapolation of monitoring results obtained at a trial instrumented section. Monitoring results are here compared with model predictions relying mostly on results from seismic dilatometer tests (SDMT). Parameters deduced from dilatometric measures are employed to characterise the different soil layers using the Hardening soil model implemented in PLAXIS. Then, the extra information provided by the seismic data is fed into a nonlinear elastic-plastic model, using a well known stiffness-strain degradation curve. Results from both simulations are here compared with the monitoring data.