Abstract: The bearing cage is a fault-prone component of the motor. Due to the small contact force between the cage and other bearing components, the impulsive vibration signal resulted from the cage fault is very weak, and it can be detected by vibration amplitude only when the crack near the bearing pocket corners has significantly expanded or even cracked, which greatly affects operation safety. In order to solve the problem that the early weak fault of the bearing cage is difficult to identify and extract, firstly, a new kind of method for selecting the best frequency bands based on the local maximum value of the kurtosis surface is proposed. Compared with other traditional methods, the proposed method could effectively find the optimum frequency bands which contain most fault characteristics in strong background noise. Moreover, a bearing fault saliency calculation method is proposed to evaluate the envelope spectra of the selected optimal bands, enabling automatic cage fault identification. After that, two sets of cage fault experiments verify the method's diagnostic effectiveness, followed by a long-term acquisition experiment to assess its robustness.

Key words: motor bearing, cage fault, kurtosis surface, early fault diagnosis