Abstract: The problems of aviation safety caused by arc fault are paid much attention by people. Due to the narrow internal space of the aircraft, the characteristics of arc fault are not obvious and they are easily interfered by aviation flight conditions, which makes detection difficult. In this paper, the collected arc current signal was regarded as a one-dimensional time series. Starting from the chaotic characteristics of the system, the phase space reconstruction technology was used to introduce the arc current signal into the high-dimensional space, and the geometric and attribute characteristics of the phase plane attractor were analyzed. The variation rules of the four characteristic quantities of center distance, radius vector offset, correlation dimension and Kolmogorov entropy were analyzed before and after arc fault. Arc fault detection technology based on principal component analysis (PCA) was used to reduce the dimensionality of the feature matrix. At the same time, the monitoring index of the squared prediction error (SPE) of the PCA and its control limit were given under each experimental load. By comparing the SPE value before and after the arc fault with its control limit, the aviation AC arc fault detection was realized. The analysis results show that the four characteristic quantities fully reflect the difference of randomness and chaos of the system before and after the occurrence of arc faults from different angles. There is no need to manually set the threshold, which can realize the unsupervised online identification of arc fault. Finally, the classification methods of different types of loads were given by principal component analysis. In order to realize the effective classification of loads while realizing the arc fault detection in the future, the arc faults can be judged and dealt with more specifically.