In order to enhance the safety and intelligence in the tunnelling of hard rock tunnel boring machines (TBM), this paper proposed a prediction model (BiLSTM-SVR model) that combined the bidirectional long short-term memory (BiLSTM) neural network with the support vector regression (SVR) algorithm, and the proposed model was capable of concurrently predicting surrounding rock grades and TBM tunnelling parameters based on the TBM tunnelling data. Example verification results show that the BiLSTM-SVR model demonstrates high accuracy in predicting surrounding rock grades, with the root mean square error (RMSE) and the mean absolute percentage error (MAPE) both being less than 0.026 5 and 0.95%, respectively. In the prediction of tunnelling parameters using the BiLSTM-SVR model, the accuracy in predicting thrust and torque is the highest, while the accuracy in predicting net excavation speed and excavation efficiency is the lowest. The proposed model exhibits significantly improved accuracy in predicting tunnelling parameters compared to the BiLSTM model and the SVR model individually, thereby combining single model effectively enhances the accuracy and robustness of model prediction.