Abstract: During the excavation process of tunnel foundation pits, traditional monitoring methods face difficulties in handling data loss and inaccurate trend prediction in complex operating environments, making it difficult to meet the high-precision requirements of engineering safety assessment. Therefore, this study integrates deep learning technology with intelligent monitoring systems to propose a safety assessment method for tunnel excavation. By integrating long-term measured data to construct a feature sample library and utilizing the high-dimensional nonlinear data processing advantages of LSTM algorithm, a dynamic monitoring and early warning model is established. Build time-series prediction models for two different working conditions: excavation period (dynamic construction stage) and stable period (structural equilibrium stage), to achieve monitoring data prediction and repair functions. The research results show that the model exhibits high prediction accuracy in both operating conditions, and the overall prediction performance of the three stable point data is good, with an overall prediction error of less than 2 mm. This effectively solves the problem of data loss processing and trend prediction in complex working environments using traditional monitoring methods. The safety assessment method for tunnel excavation proposed in this study has good robustness and significantly improves the data parsing ability of the monitoring system. It provides a new and effective method for tunnel engineering safety assessment and has practical guidance value for similar projects.