The hydraulic system of civil aircraft is one of the key systems to realize aircraft maneuvering and ensure the safe operation of aircraft. For hydraulic system fault diagnosis, researchers have carried out large number of studies, and the constructed research model can effectively identify a single fault type. With the increasing complexity of the aircraft executive function, the compound and hidden faults gradually increase, resulting in mixed fault diagnosis characteristics, which caused it is difficult to distinguish specific fault combinations. To address this dilemma, this paper proposes a composite fault diagnosis method based on Bayesian network and hydraulic system working logic, constructs a transfer model between fault characterization parameters and hydraulic system faults based on the system working principle, and proposes a composite fault diagnosis model based on the actual data of single fault, the virtual data output from the working logic model and the combination of Bayesian network. The effectiveness of the method is verified using QAR data from actual operations, and the results show that the method is extremely effective for compound fault diagnosis of hydraulic systems.