Tunnel engineering is one of the typical megaprojects given its long construction period, high construction costs and potential risks. Tunnel boring machines (TBMs) are widely used in tunnel engineering to improve work efficiency and safety. During the tunneling process, large amount of monitoring data has been recorded by TBMs to ensure construction safety. Analysis of the massive real-time monitoring data still lacks sufficiently effective methods and needs to be done manually in many cases, which brings potential dangers to construction safety. This project proposes a hybrid data mining (DM) approach to process the real-time monitoring data from TBM automatically. Three different DM techniques are combined to improve mining process and support safety management process. In order to provide people with the experience required for on-site abnormal judgment, association rule algorithm is carried out to extract relationships among TBM parameters. To supplement the formation information required for construction decision-making process, a decision tree model is developed to classify formation data. Finally, the rate of penetration (ROP) is evaluated by neural network models to find abnormal data and give early warning. The proposed method was applied to a tunnel project in China and the application results verified that the method provided an accurate and efficient way to analyze real-time TBM monitoring data for safety management during TBM construction.