An Adaptive Optimization Feature Extraction Method Based on Firefly Algorithm for Motor Bearing Fault Diagnosis

In motor bearing fault diagnosis, extracting effective information from signals is a critical prerequisite, and the complete ensemble empirical mode decomposition (CEEMD) can be used to process signals. This paper proposes an adaptive optimization Gaussian white noise (GWN) selection method based on the firefly algorithm (FA) for bearing fault diagnosis, which avoids the mode mixing phenomenon. The index of orthogonality is used to represent the degree of mode mixing phenomenon, and the value of GWN amplitude is optimized by FA to reduce the degree of the mode mixing phenomenon in the intrinsic mode functions (IMFs). The fault feature information will be more evident in the IMFs when the mode aliasing is suppressed. The experimental results show that the appropriate amplitude of the GWN can be obtained by the optimization method for different fault signals and adaptively control the mode mixing phenomenon. In addition, the spectrum diagram of the IMfs shows the fault feature information clearly.