Abstract: Driving efficiency of a TBM is closely related to the characteristics of the surrounding rock and machine performance, which are the two factors that need to be considered for predicting the TBM advance rate. Because of the uncertainty of geological factors, the statistical distribution of UCS and RQD are simulated using a normal distribution and an exponential distribution, and the corresponding random input parameters are generated with the Monte Carlo algorithm. When considering the factors of machine performance, the differences caused by the influence of various TBM performance factors are eliminated with the ratio of the net thrust and cutterhead diameter.Based on the inputs of both the surrounding rock and machine performance parameters, a BP neural network model to predict TBM penetration is set up in Matlab. Practical cases verify that the predicted results are close to the measured ones.