In this paper, we propose a novel fault diagnosis method for aircraft power supply system. We constructed the Dynamic Fault Tree (DFT) of the Constant Speed Constant Frequency Alternation Current (CSCF-AC) power supply system based on the Priority Dynamic Fault Tree (PDFT). And by adding a new temporal operator BEFORE (BF) to the usual Boolean operators, we derived the structure function of the top event. Then two laws were provided to reduce this structure function to a sum-of-product canonical form, where each product term defined a minimal cut set (MCS) or a minimal cut sequence set (MCSS) of this fault tree. This approach provides a complete qualitative description of the system without resorting to the corresponding Markov model. Then a 3-layer network diagnosis system was designed and realized for the fault diagnosis of aircraft power supply, including onboard layer, ground maintenance layer and remote expert layer. This 3-layer system enhances the flow of complex fault data into higher expert layer to realize the advanced diagnosis.