The invention discloses a fault diagnosis method of a fuel regulator of an aero-engine. The method comprises the following steps: on the basis of a part-level modeling method, a high precision model of a fuel regulator is established; on the basis of the conventional model, fault parameters are introduced by means of soft faults to simulate various typical faults that may occur in a real fuel regulator; a parameter that can reflect the system state is selected from a measurable output value of the fuel regulator and parameter feature values in various fault modes are extracted; sample sets ofdifferent fault modes are established and the sample sets are divided into training samples and testing samples; the samples are trained and tested by using a BP neural network as a classifier so as to realize the fault diagnosis of the fuel regulator. According to the invention, problems of many parameters for collection, tedious operations, and low diagnosis accuracy according to the existing fuel regulator fault diagnosis technology are solved. The fault diagnosis method can be applied to the field of fault diagnosis of aero engine fuel regulators or the fault diagnosis field of other hydraulic systems.