The sufficiency and integrity of dynamic stress spectrum will affect the accuracy of the analysis result. It is necessary to extrapolate them from the short term to the long term while retaining the probability distribution characteristics. The stress spectrum extrapolation and fatigue life assessment of the bogie frame were studied, and a test point was taken as a case study. First, the optimal bandwidth determination in two-dimensional KDE (kernel density estimation) was studied. Second, the gray incidence degree analysis method was proposed to evaluate the proximity-incidence degree and similarityincidence degree of the rain-flow matrices extrapolated before and after.Finally, a fatigue life assessment method, with a multi-sample dynamic stress spectrum extrapolated based on KDE, was proposed. Compared with the linear extrapolation, the relative error of fatigue life assessment result based on KDE extrapolation is −4.17%, which meets the accuracy requirements of engineering. It is more beneficial to ensure the vehicles' safe operation.