Joints of shield tunnels are the weak but key elements of the whole structure. Considering the resistance capacity to bending moment, shear force and seepage, assessment of the vulnerability of shield tunnel segment joints was conducted and a model of vulnerability evaluation of segment joints subjected to inner forces was established under the consideration of the health state of joints. The stress of segment concrete and bolts, and seepage were taken as the assessment criteria of the model. Then, an analytical joint model was established to analyze the mechanical response of joints, and verified by the 3D refined numerical simulation. Based on a large number of learning cases offered by Monte Carlo simulation analysis, Bayesian networks were employed to evaluate the vulnerability of segment joints. Based on insitu monitoring data, the back analysis of health state of shield tunnel segment joints and the updated prediction of the vulnerability of segment joints are applied to help the operation and maintenance of shield tunnels.