Abstract: The core of TBM tunnel construction is penetration operation and the reasonable penetration velocity is of great significance to improve construction efficiency and reduce construction risk. TBM penetration velocity is affected by many factors, among which there are complex nonlinear relationships, so a penetration velocity prediction mode for TBM based on PCA-RVM was proposed in this paper. Principal component analysis(PCA) was used to reduce the five influencing factors(uniaxial compressive strength,Brazilian tensile strength,distance between plane of weakness,punch slope index,alpha angle between tunnel axis and the planes of weakness) into three independent principal component variables,and the relevance vector machine(RVM) was used to establish the nonlinear mapping relationship between three independent principal component variables and the penetration velocity of TBM. Then the penetration velocity for samples only with influencing factors can be accurately predicted. The model was applied in an actual project and compared with the results by relevance vector machine and BP neural network. It is shown that the PCA-RVM prediction model has advantages of high prediction accuracy,good fitting degree,small discreteness and well suiting for small samples.