This paper analyzes the residual sequence between the output of the fitting model and the real flight parameter data, extracts time-domain eigenvalues that can represent the characteristics of the residual sequence through the fault injection method, reduces the dimension and makes mathematical statistics of time-domain eigenvalues of the residual sequence by using Principal Component Analysis, realizes the fault identification of aircraft fly-by-wire control system, and finally classifies time-domain eigenvalues of the aircraft residual sequence by using Extreme Learning Machine, realize the fault diagnosis of aircraft fly-by-wire control system.