Abstract: The control of the excavation posture of shield tunneling machines is critical to ensuring both tunnel construction quality and the protection of the surrounding environment.Inadequate posture control can result in a range of engineering challenges.Currently,posture control in shield tunneling is largely based on engineering experience,with no precise or feasible method available for verifying the excavation posture.To address this gap,this study proposes a method for predicting the attitude of slurry balance shield machines using a combined Transformer- LSTM model.By incorporating 26 different construction parameters from practical engineering applications,the study employs correlation analysis and hierarchical clustering to select input features for the model,which is then used to predict the excavation posture of the shield machine.The results demonstrate that the Mean Absolute Error (MAE) for vertical and horizontal deviations in excavation posture,as predicted by the Transformer-LSTM model,are 0.054 and 0.061,respectively,while the Root Mean Squared Error (RMSE) values are 0.064 and 0.085.Additionally,the correlation coefficients are 0.883 and 0.913,respectively.These findings show that the Transformer-LSTM model significantly outperforms the standalone LSTM model,making it a promising tool for assisting in the verification of shield machine excavation posture.