Abstract: Due to the high similarity of the fault characteristics of aviation transformer rectifier diodes in different fault modes, it is difficult to distinguish the fault characteristics. To solve the problem, a fault diagnosis method was proposed based on stacked denoising auto encoder (SDAE) combined with particle swarm optimization support vector machine (PSOSVM). Firstly, a simulation model of aviation transformer rectifier was built to get the fault data from simulation of different fault modes. Then the SDAE method was used to extract the fault features from high-dimensional fault signals and establish fault feature sets. Finally, PSOSVM method was used to diagnose fault and to compare the effectiveness with common fault diagnosis methods. The fault diagnosis results show that the accuracy of the SDAE-PSOSVM method can reach up to 96% and functionally extract the features of high-dimensional fault data signals to improve the discrimination between different fault modes.