Planetary gearbox is easy to fail in the operation of equipment, and its vibration signals are mostly coupling, multi-transfer path and nonlinear. In order to reduce the cost of sensors placement and improve the accuracy of fault identification for monitoring system, in this paper, an optimization method for measuring points of planetary gearbox was presented based on the combination of local mean decomposition (LMD) information entropy and correlation analysis, and it was applied to the fault diagnosis of planetary gearbox. Firstly, the vibration test signals of five working conditions were decomposed by LMD algorithm. Then, the PF components containing the main fault information were screened out by the correlation coefficients between the decomposed components and the original data, and their information entropy features were extracted and constructed into sample feature vectors; Finally, the correlation analysis and calculation of the information entropy feature vectors of different measuring points in each working condition were carried out. The correlation indexes among the measuring points were statistically analyzed, and six measuring points were selected from original eleven measuring points; at last, the validity of the measuring points was verified by the diagnostic results of ALNAFSA-BP network for planetary gearbox