Abstract：Compared with bearing and fixed axle gear box, it is more difficult to extract the fault characteristics of the planetary gearbox, and the traditional method is not good for the diagnosis of planetary gearbox. Therefore, in this paper, a new method of fault feature extraction based on wavelet packet entropy and RMS is proposed to solve the nonlinear and complex characteristics of fault vibration signal of planetary gearbox. Firstly, the original signal is intercepted by a sample of the same length, and then the wavelet packet transform is used to decompose the sample signal, and the sample entropy and root mean square value of each frequency band are calculated and normalized. The normalized parameter is used as the weighted average weight to calculate the weighted average sample entropy and root mean square value. Finally, two parameters are used to get new parameters. Based on the results of fault diagnosis and anti noise test, the new feature extraction method can increase the differentiation degree of the fault characteristics of planetary gearbox and has good stability, at the same time, the new parameters have a certain anti noise.