Abstract: In order to accurately evaluate the damage degree of planetary normal and to improve the evaluation guidance capability for equipment maintenance, a translation-torsion dynamic model was established for planetary gear train. Analyzing the variation of spectrum amplitude and crack degree of planetary gear train in dynamic simulation signal, a new crack depth evaluation feature was proposed based on the Ratio of Cumulative amplitude of Multiple Fault frequency to Amplitude of Meshing frequency (RCMFAM). The results show that, compared with the typical time-domain and frequency-domain features, the proposed feature can offer a better correlation and strong consistency with the crack damage degree. In the validating experiments with normal, 50% crack damage and broken tooth gear, the evaluation results are basically consistent with the actual crack depth, verifying the effectiveness of the proposed features.