The recorded traffic data show that traffic loads have been increasing. In this study, based on the long-term health monitoring data collected from Anhui Province, the annual average hourly traffic (AAHT) is defined with taking into account the periodical and seasonal change of traffic volumes, and furthermore, an autoregressive moving average model (SARIMA) is established to simulate truck loads in the future. At the same time, several truck-load models are developed with the statistics of the key parameters of truck data, and then loaded one by one on the finite element model of a T-bridge to calculate the fatigue damage induced by the non-stationary increases of truck traffic. The results show that the AAHT-based SARIMA model is accurate and efficient for predicting traffic loads, and the non-stationary increase of traffic loads will significantly jeopardize bridge structures. The fatigue damage with considering the non-stationary increase is about 1.7 times of the fatigue damage without considering the non-stationary increase.