Real-Time Prediction of Operating Parameter of TBM during Tunneling

With the increasing use of the tunnel boring machine (TBM), attempts have been made to predict TBM operating parameters. Prediction of operating parameters is still an important step in the adaptability of the TBM for the future. In this study, we employ a walk forward (WF) prediction method based on ARIMAX, which can consider time-varying features and geological conditions. This method is applied to two different TBM projects to evaluate its performance, and is then compared with WF based on ordinary least squares (OLS). The simulation results show that the ARIMAX predictor outperforms the OLS predictor in both projects. For practical applications, an additional analysis is carried out according to the real-time prediction distance. The results show that time series-based ARIMAX provides meaningful results in 8 rings (11 m) or less of real-time prediction distance. The WF based on ARIMAX can provide reasonable TBM operating conditions with time-varying data and can be utilized in decision-making to improve excavation performance.