Application of an improved EEMD method in bearing fault diagnosis of induction motors

  Objective  In order to overcome the disadvantages of the traditional ensemble empirical mode decomposition (EEMD) method in selecting parameters (integration time and white noise amplitude coefficient) based on experience, and reduce the cost of calculation time, a fast ensemble empirical mode decomposition (FEEMD) method is used to extract the characteristic frequency.

  Method  By changing the distribution density of the added white noise, different signal envelopes can be obtained. Furthermore, we can identify the optimal envelope by finding the optimal search window width of the moving mean filter, thereby avoiding the defect of EEMD selecting parameters by experience. At the same time, after the abnormal component in the signal is decomposed, the residual component can be decomposed by EMD to further save the calculation time cost. Finally, the method is combined with Hilbert envelope demodulation technology and applied to the fault characteristic frequency diagnosis of the bearing inner ring of an asynchronous motor.

  Results  As the results show, compared with the traditional EEMD method, FEEMD can extract the fault frequency more efficiently.

  Conclusion  FEEMD overcomes the disadvantages of the traditional EEMD method in selecting parameters based on experience and shortens the calculation time. As such, it can be effectively applied in bearing fault frequency extraction experiments.