A planetary gearbox fault signal classification method based on DWT-LPP comprises the steps of S1) collecting vibration signals; S2) performing discrete wavelet transformation on the collected vibration signals, and expanding one-dimensional signals to multi-dimensional signal matrixes with various frequency domain scales; S3) carrying out dimension reduction on the multi-dimensional signal matrixby adopting a local preserving mapping algorithm to obtain a mapping result of preserving signal local manifold structure information. The planetary gearbox fault signal classification method based on the DWT-LPP has the advantages that fault vibration signals are enhanced, noise signals are restrained, and different fault types can be effectively distinguished.