As a key component of the main reducer in aviation helicopters, the bending fatigue strength of spiral bevel gears has attracted much attention. However, due to high experimental costs and long cycles, there is a lack of relevant data. This article proposes a reliable and efficient method for evaluating the bending fatigue strength and predicting the fatigue life of small sample bevel gears. Firstly, a pair of spiral bevel gear test samples were designed, and 3D modeling, assembly, and simulated rolling inspection were completed. Based on transient dynamics and finite element fatigue life simulation, the bending stress variation history of the tooth root was obtained. Then, a closed power flow bevel gear durability test rig was built, and S-N fatigue life data was obtained through fatigue testing. To address the issue of limited data, the GM (1,1) model is introduced, and through residual correction and logarithmic processing, the prediction accuracy is improved to 0.71%. To improve data reliability, the S-N data samples were expanded by integrating finite element simulation, GM (1,1) model prediction, and experimental statistics. After fitting and testing, the two-parameter Weibull distribution has a better fitting effect, and the R-S-N curve of the material is determined based on this distribution.