High-performance servo control strategies often require accurate and reliable mathematical model support. The mathematical model characterized by inductance needs to fit the actual electromagnetic saturation process in the form of dynamic inductance under short-term overload conditions of the motor. Aiming at this problem, this paper uses the full flux model to characterize the electromagnetic saturation characteristics of the motor under short-term overload, and constructs a fully connected neural network to realize the identification of flux linkage. Firstly, the defects of the traditional inductance model under overload are analyzed, and the mapping relationship between flux linkage and current is proved. Secondly, the fitting extraction method of variable torque flux linkage data is constructed, and the identification method of flux linkage-current mapping surface is established by fully connected neural network. Finally, the flux-current data obtained by the finite element parameter scanning and the flux-current data collected by the experiment are used to identify the mapping surface, and the characteristics of the motor overload electromagnetic saturation are analyzed by the experimental results.