Based on the average fatigue life of LZ50 axle steel specimens without surface rolling, five maintenance times were determined. Accordingly, five groups of specimens were turned and rolled at above maintenance times and were fatigued using a replica technique. The results show that the crack growth rate is much lower than that before rolling at a given dominant short crack size. The effective short crack density of all specimens decreases significantly after maintenance. However, with the postponement of surface maintenance, the highest effective short crack density for the five studied groups of specimens increases continuously, while the average fatigue life decreases gradually. A maintenance time effect function is presented to refine a short crack growth model described previously. The refined model can include a significant effect of the maintenance time on short crack growth and predict its patterns at different maintenance times with exiting test results for LZ50 axle steel.