Abstract: Aviation series arc fault has strong concealment, which leads to the difficulty of detection, and seriously endangers the safety of aviation flight. This problem has been paid more and more attention. In view of the limitation of the current time-frequency method for series arc fault diagnosis, the generalized S-transform was proposed to analyze the experimental current signal and extract the characteristic quantity. According to the standard, a series arc fault simulation device and crosstalk immunity experiment were built, and the current signal in the circuit was collected. The experimental data was intercepted by the signal selection box for generalized S-transform, which improved the accuracy of arc fault identification. According to the change of frequency spectrum after the fault occurs, the root mean square value and energy of 2kHz component were extracted as the characteristic quantity. Compared with the analysis results of wavelet transform method, the results show that the method has higher recognition accuracy in Aviation series arc fault diagnosis, and crosstalk immunity will not be misjudged. Finally, the extracted feature vectors were constructed and then input into the support vector machine optimized by particle swarm optimization for recognition. The results show that the recognition accuracy reaches 98.75%, which is further improved compared with the single feature value. It provides a reliable reference for the development of aviation arc fault circuit breaker.