As the industry upgrades, the reliability demand has gradually increased, the fault detection of motor has become the current research hotspot. Since the available detecting signals of the motor are few, the load condition is complex, and there exists signal processing error, the characteristics of broken rotor bar of squirrel-cage motor is difficult to detect. Regarding the broken rotor bar (bad bar) as the research object, finite element simulation and modeling of broken rotor bar are carried out, a variety of signal processing methods are studied comparatively, and a detection method with high accuracy and anti-interference ability is proposed. It can achieve the goal of distinguishing the broken rotor bar (bad bar) type and predicting the fault degree.