AbstractStochastic traffic flow, as a type of repeated load, can cause serious high-cycle fatigue damage to bridges. In addition, the rough simulation of stochastic traffic flow and inappropriate analysis method of fatigue stresses cause the fatigue evaluation results to deviate from reality. To overcome this challenge, a probabilistic fatigue valuation method is proposed based on an elaborate simulation of the stochastic traffic flow and field-measured strain influence line. By selecting vehicle load features affecting the bridge structural fatigue as clustering parameters, the two-step clustering (TSC) method is applied to distinguish the different traffic states with the clustering numbers to be determined objectively. On this basis, the elaborate stochastic traffic flow is simulated by random sampling of vehicle feature probabilistic models for each traffic state. Subsequently, the bridge strain influence line, which is identified through synchronous monitoring of strain and vehicle positions, is loaded by the simulated traffic loads to obtain the stress history instead of the traditional finite-element model (FEM). Finally, the structural fatigue life can be probabilistically predicted through a Monte Carlo simulation. The proposed method was verified to be effective through a case study of a long-span suspension bridge. It can be concluded that distinguishing the different traffic states can improve the rationality of stochastic vehicle load simulation, and a more reasonable prediction of the vehicle-induced bridge fatigue damage can be obtained through the influence line loaded by stochastic vehicle loads.