Abstract: To achieve the early trouble shooting of the rolling bearings in tobacco primary processing equipment operating under low-speed, light-load and noisy conditions, a fractional order stochastic resonance method based on noise utilization mechanism was proposed. The method enhances the energy of trouble features by converting the noise energy in the bearing vibration signal and the vibration energy of other normal components into the weak trouble features. In the meantime, the time-delay dependence of fractional derivatives and the memory characteristics of historical observation information in the process of feedback are utilized to boost the enhancement effect of stochastic resonance. The proposed method was applied to achieve early trouble shooting of rolling bearings in primary processing equipment, and its effects on engineering application effects were verified. The results showed that: 1) The original vibration signal of rolling bearings was noisy with a low ratio of signal to noise, and therefore the information about trouble features was hard to be observed from its frequency spectrum and envelope spectrum. 2) When the infogram method and the underdamping stochastic resonance bearing characteristic enhancement method failed in trouble shooting, almost all of the vibration energy in the enhanced result was presented at the trouble feature frequencies when the original vibration signal was processed by the proposed method, indicating trouble features of bearing. The analysis result was consistent with the actual state. 3) After applying this method to ten machines in four processing lines in the primary processing department of Ningbo Cigarette Factory for one year, a total of five bearing troubles missed by regular manual patrol inspections/spot checks were detected. This method provides support for improving the effectiveness of tobacco primary processing equipment maintenance.