Abstract: Traditional prediction model of TBM tunneling focuses on using a single factor to predict the TBM tunneling performance, and the predicted results are often quite different from the actual construction.In this paper, based on a tunnel project in Xinjiang, we proposed a TBM tunneling prediction model which can consider many factors and conform to the actual engineering.Firstly, the penetration index FPI was taken as an evaluation index of TBM′s excavability, and the correlation between the in-situ measured rock mass integrity indexKv,uniaxial compressive strength UCS and FPI was analyzed.Based on the correlation analysis, a multiple regression model was established withKvand UCS as input parameters and FPI as dependent variable.Then, a fitting relation between FPI and cutter speed RPM and driving speed measured on site was established, by which the cutter speed and driving speed in actual construction projects can be predicted.Finally, on the basis of FPI prediction model, K-means clustering method was used to classify the TBM′s tunneling grade, and reasonable tunneling parameters were estimated according to the tunneling grade.The results show that the TBM tunneling model established in this paper can well reflect the real state of TBM construction, and can provide a reference for project schedule prediction and cost control in the TBM construction process with similar formation conditions.